Digest / June 12, 2026
Claude Code Issues Digest — June 12, 2026
395 new issues filed ·
338 resolved in anthropics/claude-code (UTC day).
New issues (395)
- [Bug] Model incorrectly flags benign conversation and falls back to Opus
- [Bug] Computer use requests not sent through Remote Control approval interface
- [Bug] Adversarial code review request rejected by safety filters
- [Bug] Agent infinite loop during execution
- [BUG] Cowork agentic mode: plugin OAuth helper loses flow state between turns — authorization codes issued but never exchanged ("No OAuth flow is in progress")
- [Bug] Thinking state never completes, no progress indicators or token/elapsed time display
- Remote Control: hook-set sessionTitle never syncs to the RC session name (web/iOS keep hostname auto-names); /rename does sync — per-channel matrix
- [BUG] Answer to a queued mid-turn user message is written between tool calls and never displayed; model later insists it answered
- [FEATURE] Allow the setting of the start up color.
- [FEATURE] Feature request: Live file references between threads ("xrefs" for Claude)
- [Bug Report] Unable to process empty bug report submission
- Feature: Cross-project session transfer / switch-project command
- [BUG] Citation tags incorrectly insert source information in the middle of quoted text
- [BUG] v2.1.172 MCP OAuth regression: reused client registration causes registration↔authorization scope mismatch → invalid_scope ("not allowed to request scope 'openid'") on fresh server auth (Clerk)
- [BUG] [macOS] OAuth login fatally errors instead of falling back to manual token entry when loopback callback can't bind (sandbox-runtime); does not repro on Linux
- [Bug] Unexpected High API Cost Usage
- [Fable 5] Harness forced model-switch back to Opus 4.8: examples
- [Bug] Model switching during execution breaks security hardening workflow
- [BUG] /feedback sends Git repository + SSH commands with customer personal data
- Auto-mode permission classifier denies plan-approved and explicitly user-directed gh commands with factually incorrect rationales
- [BUG] Claude Code Web Audio Mode retains previous message content after send
- [Bug] False-positive safety classifier triggers model downgrade during legitimate legal/security work
- [FEATURE] Generate session titles in the user's conversation language (e.g. Japanese)
- Bash tool: shell cwd resets to project root between tool calls, breaks worktree workflows
- Desktop preview: preview_click lands off-target after preview_resize (coordinates not adjusted for emulation scale-to-fit)
- [Bug] False positive model-switch trigger in 3D graphics discussion
- [BUG] + [RFC] Tasks run forever: 7-layer architectural root cause + permanent fix proposal (PALMS)
- [BUG] Bedrock: --thinking-display summarized is ignored — thinking blocks arrive with empty text (identical command works on first-party API)
- Subagents return fully hallucinated results with zero tool calls; leaked tool-call XML in text; two fabricated 'prompt injection detected' reports
- [BUG]
- [Bug] Claude Fable 5 flagged legitimate security discussion and downgraded to Opus
- [Feature Request] Support for security audit analysis across repositories
- [Bug] Anthropic API Error: Content Moderation Rejection for Defense Use Case
- Sandboxed Bash commands get a spurious backslash before `!` (shell-quote re-escapes the eval wrapper)
- [FEATURE] URGENT: Nigeria pricing arbitrage by bulk gift card resellers
- [Bug] Anthropic API Error: Content flagged as problematic for model
- [Bug] Anthropic API Error: Content flagged as problematic for model
- [BUG] Sandbox reports that github.com DNS resolution fails
- [DOCS] [Fullscreen] New `wheelScrollAccelerationEnabled` setting added in v2.1.174 is undocumented in `fullscreen.md` and the settings table
- [DOCS] [Model config] `/model` picker now exposes the Default-resolved family as its own row, but `model-config.md` does not describe the picker layout after v2.1.174
- [DOCS] [Amazon Bedrock] GovCloud (`us-gov-*`) regions need `us-gov.` inference-profile prefix, not the `global` prefix; `amazon-bedrock.md` does not document GovCloud at all
- [DOCS] [Agent view] Background-session supervisor no longer inherits the parent shell's `ANTHROPIC_*` env on dispatch, but `agent-view.md` never said the supervisor inherited from the starting shell in the first place
- [DOCS] [Advisor] `/advisor` picker pre-selects the last saved advisor model, but `advisor.md` does not say the picker refuses to pre-select a model blocked by `availableModels`
- [DOCS] [Workflows] Workflow `agent()` subagents were missing per-agent attribution headers, but `workflows.md` does not document the per-agent attribution layer at all
- [DOCS] [VS Code] Account & usage dialog `/usage` now shows per-skill/agent/plugin/MCP cache and long-context attribution, but neither `vs-code.md` nor `monitoring-usage.md` describes the new VS Code usage panel
- [DOCS] [Agent view] Pre-warmed background workers no longer fail with "Could not resolve authentication method" when claimed after sitting idle, but the pre-warmed worker dispatch path is still not documented in `agent-view.md`
- Expose session usage-limit state to the model and hooks, with a configurable threshold and automatic post-pause resume
- [BUG] Windows: case-sensitive `.claude.json` project-key lookup silently drops MCP servers across CLI/VS Code — recurring bug, repeatedly auto-closed unfixed
- [BUG] Claude Max 5x gift subscription redeemed but subscription end date not extended — billing system failed to apply gift
- False-positive Usage Policy block on benign procedural-animation prompt
- [Bug] Anthropic API Error: Server rate limiting on requests
- [FEATURE] Expose usage-limit reporting and add hook points for graceful shutdown and resume
- hookSpecificOutput.sessionTitle does not propagate to desktop app sidebar / remote session title (unlike /rename)
- WebSearch tool broken: internal model claude-haiku-4-5@20251001 not found
- I appreciate you reaching out, but I need to clarify a few things: 1. **I'm Claude (the AI assistant)**, not Claude Code. Claude Code is a separate CLI tool for coding tasks. 2. **I don't switch models** - I'm Claude, and I remain Claude throughout our c
- [FEATURE] Native Linux desktop app
- Telegram channels plugin keeps consuming getUpdates while the CLI refuses channel registration — permanent inbound message loss
- Channels: inbound message left as stale composer render with empty input buffer (lost-Enter ghost) — never submits, retry re-ghosts
- Fable 5: assistant text emitted before tool calls is silently dropped from terminal output
- [Bug Report] Unable to generate issue title - no bug report content provided
- [BUG] Tab-indented code renders without indentation in Write/create-file preview (file content is correct)
- [Bug] Anthropic API Error: Invalid model combination for tool advisor
- [BUG] Streaming partial-JSON parser silently produces empty MCP tool arguments (accumulator shear)
- `API Error: The socket connection was closed unexpectedly` — packet captures show server-initiated FIN mid-stream (10 incidents, requestIds included)
- [Bug] Anthropic API returning safety filters on legitimate jiu jitsu biometric content
- [Bug] Agent workflow resumes rerun completed tasks instead of resuming from checkpoint
- [FEATURE] Runtime-enforced verification gates for Skills
- File viewer: add preview (eye) button for .html files, same as Markdown
- [Bug] Safety classifier false positives on defensive security engineering and governance work
- [FEATURE] Team/org-level /stats view (UI + terminal) for admins
- [FEATURE] Headless / exportable /stats (--json and image export)
- [Bug] Anthropic API Error: Inappropriate Content Flag on Wellness Genome Chat Application
- [FEATURE] Maintain scroll position and keep the composer in view when typing while scrolled up
- New chat submission steals focus to Chrome window (macOS Space switch) when Claude in Chrome is paired
- [Bug] False positive security guardrail detection
- Claude in Chrome (Linux): CLI never attempts bridge socket connection — all tools return 'Browser extension is not connected' after exactly 10s, while extension↔native-host channel is verified working
- Model picker state diverges between desktop and mobile clients on a remote-controlled code session
- **[BUG] Cowork locks up after ~5 min — EventEmitter memory leak (MaxListenersExceededWarning) accumulates per session, Electron renderer freezes**
- Claude desktop app leaks /dev/ptmx master fds, exhausting the OS pty pool (forkpty: Device not configured)
- [DOCS] [Model config] `enforceAvailableModels` managed setting added in v2.1.175 is undocumented in `model-config.md`, the settings table, and the Managed-only settings list
- [BUG] Diff rendering: bold/code text invisible on red (deleted) background lines
- Background agent re-runs after completion: duplicate task-notifications under same task ID with uncommissioned work (2.1.174)
- Feature request: setting to hide individual built-in slash commands from the menu
- Claude desktop app leaks PTYs, exhausting system-wide kern.tty.ptmx_max (forkpty: Device not configured)
- [Bug] Model emits malformed tool calls with corrupted invocation tokens
- The domain desktop-releases.claude.ai fails DNS resolution globally (tested against 8.8.8.8 and 1.1.1.1), preventing the Cowork VM bundle from ever downloading. CoworkVMService cannot start because vm_bundles never gets created
- Feedback: Preview tool needs device/viewport presets, better stability, and UX polish
- [Bug] Opus 4.8 (xhigh) fabricated a full make build + pkg artifact, lied about file path/size/commit hash, then doubled down through two user errors
- [BUG] macOS desktop hangs on launch after auto-update — gray screen / dock bounces forever; orphaned crashpad handlers survive quit and block early init
- Telegram channel plugin: bun server self-kills / fails to spawn on Windows (v2.1.174)
- [BUG] Claude Code modifies code without being asked
- [BUG] Resumed background subagent/teammate loses its spawn-time model override and reverts to parent model
- [Feature Request] Add safety override or allowlist for inference-focused repositories
- [FEATURE] Claude's Knowledge Gap: Memory System Implementation Details
- [FEATURE] Expose rate limit / 5-hour usage data for status line and external tooling
- [DOCS] Outdated homebrew installation script
- [Bug] Unnecessary security warning for local workspace file changes
- [BUG] MSIX + Smart App Control: Windows-MCP blocked (os error 4551), and Directory extension install silently fails ("installDxtFromDirectory: reply was never sent") on 1.12603.1
- [BUG] — Claude Code Windows v2.1.175 launches as Bun instead of Claude
- [Bug] Model selection reverts to Opus 4.8 instead of respecting configured model
- Claude Desktop scheduled tasks: completed sessions never exit — unbounded memory accumulation (~0.5–0.8 GB/hr)
- preview_start: autoPort always fails on Windows/Git Bash with "port in use" for all free ports
- preview_start breaks on Windows/Git Bash — autoPort returns "in use" for all free ports (regression)
- [BUG] Preventive auto-compact and blocking-limit warning silently disabled for most recent models; desktop-app resume disarms all 1M-context sessions
- [BUG] Claude Code consumes provider deltas after cancellation
- [BUG] Kit MCP server OAuth fails with PKCE rejection despite code_challenge being present
- [BUG] Claude Code consumes provider deltas after message_stop
- [Bug] Legitimate infrastructure activity incorrectly flagged as security violation
- [BUG] Claude Desktop archived an active Remote Control session overnight without user action (history inaccessible, no undo)
- [BUG] Claude Code does not bound repeated identical tool calls
- [BUG] Claude Code returns tool-result state after cancellation
- [FEATURE] Hook output primitive to require a tool call (`requiredToolCalls`) — hooks can veto and inform, but cannot initiate
- [FEATURE] Machine-readable hook I/O schemas shipped with the binary (`claude hooks schema --json`)
- new sessions into specific groups
- Feature request: Generate session titles in the user's conversation language
- [Bug] Claude Code auto-switches to Opus 4.8 with Fable 5 model selection
- [BUG] Cowork VM leaks NtFC (NTFS FCB) NonPaged Pool at 12-16 GB/h; two full RAM-exhaustion cycles in one evening, reboot required (re #45921)
- Regression: Arrow Up/Down no longer moves cursor to line start/end at boundary lines in input editor
- [BUG] /desktop on Windows spawns bare `cmd` through the where.exe PATH resolver — failed resolution cached as null for entire session; only generic 'Failed to open Claude Desktop' shown while the real error and deepLinkUrl are swallowed
- [CRITICAL] Tool call failures are destroying usability — silent errors, wasted tokens, broken workflows
- Support direct image paste from clipboard in chat interface
- [BUG] `claude://cowork/new` deeplink not working for multiple files
- [Bug Report] Incomplete Issue Submission
- [BUG] Using `claude://` deeplink does not bring Claude app to front
- [Feature Request] Improve Fable safety switches around authentication and authorization for complex system design
- [BUG] Cowork (macOS): attached image filenames not passed to Claude (works in Chat)
- [BUG] CoworkVMService crashes with "Incorrect function" (exit code 1) on Windows 11 Pro — domain-joined machine, Hyper-V enabled
- Fable model Deferred to Opus mid session, for very weak reasons in my opinion
- [Bug] Anthropic API Error: False-positive Usage Policy block on benign video-production QA task
- [Bug] Anthropic API: Fable 5 safety classifier incorrectly flags all new sessions
- subscriptionType shows "pro" instead of "max" for Max 5x subscribers
- Terminal re-renders entire output repeatedly during streaming when response contains large markdown tables
- Session archive is one-way: no unarchive tool and no Archived view in desktop
- macOS desktop app leaks PTY masters (/dev/ptmx) until system-wide PTY exhaustion (ENXIO)
- [Bug] Agent ignores user input during execution
- [BUG] Build error: duplicate createRequire imports prevent daemon startup
- [BUG]
- [BUG] Remote VM session: tool call in flight at laptop sleep phantom-denies after exactly 600s — "user doesn't want to take this action" + toolUseResult "Error: undefined", despite bypassPermissions
- [BUG] Plugins Installed UI nests only the first bundled MCP server (still reproducible in 2.1.175)
- [FEATURE] iOS App - option to launch directly into Claude code interface
- [BUG] Filesystem extension silently fails to install after Squirrel → MSIX upgrade on Windows
- [Feature Request] Differentiate security-enhancing code modifications from security-degrading ones in safeguards
- [Bug] Intent Classifier Misclassification: UI Layout Task Incorrectly Categorized as Biology/Cybersecurity
- [BUG] Slack connector: `slack_read_channel` drops `sh_room_ended` events, making huddle duration inaccessible
- Opus 4.8 fabricates entire tool executions inside extended thinking — no tool_use emitted, model believes tools ran and reports fake results
- Ultrareview (research preview): review driver dies in cloud sandbox — `CLAUDE_CODE_OAUTH_TOKEN not set`; token never injected from Claude.ai login
- Windows: path-scoped Write/Edit permission rules never match — Write tool absolutizes file_path pre-check and absolute Windows paths match no documented pattern form
- Silent data loss: app update wiped all session conversation content (IndexedDB reset without backup, no user warning)
- [FEATURE] Monthly usage summary view in console - for buget reporting and planning. Current 5-house/weekly virew are insufficient for monthly billing cycles.
- [BUG] claude_code_cost_usage_USD_total metric semantics appear inconsistent with Prometheus Counter behavior
- Feature Request: Support XDG Base Directory Specification for config/data directories
- Assistant text block silently not rendered when followed by tool calls after a mid-turn user message
- [MODEL] Opus 4.8 falsely confesses to tool-output fabrication that never happened, then spirals: subsequent self-observations reinterpreted through the false confession (JSONL-verified)
- [Bug] Performance regression: Significantly slower than Codex with repeated task misinterpretation
- Middle-click paste fires twice (on mousedown and mouseup) in VS Code integrated terminal — regression in 2.1.175
- Identifying many parallel sessions: let the model (and scripts) set/update the session name
- [Bug] Safety guard interception blocking valid operations
- [BUG] IntelliJ MCP: Connection fails with -32000 when project path contains spaces (Windows, OneDrive)
- [BUG] False auth conflict warning when using third-party ANTHROPIC_BASE_URL with both API key and /login managed key
- Three socket-close API errors in one session, each at sub-agent-return boundary (claude 2.1.144)
- [Bug] False-positive report
- [VSCode extension] Allow configuring permission mode after plan approval
- [BUG] Claude Desktop silently hangs installing any local .mcpb with a deflated entry larger than ~16 KB
- [Bug] Security scan false positive from external code review tool
- [Bug] Missing exponential backoff retry for Anthropic API rate limit errors
- Third-party marketplace with autoUpdate:true not refreshed at session start; new release not picked up across restarts (v2.1.175)
- [BUG] Images sent by agent via SendUserFile never render in Desktop app chat (local session, claude.ai webview)
- Allow forking a session as a new session from a checkpoint (branch, not in-place rewind)
- Classifier overrides user CLAUDE.md context for legitimate server admin tasks — and blocks the user from writing the fix
- bug : Remote Control: /rc loops "connecting… → disconnected", never reaches Connected (Windows 11, Pro)
- [BUG] [Cowork] Dispatch: per-tool MCP permission prompts in child task sessions are not forwarded to the mobile thread (regression)
- Add timestamps to chat messages
- [Bug] Excessive token consumption due to model guessing behavior
- [BUG] VS Code extension: edited files open in Claude's tab panel instead of editor tabs when chat has focus
- [BUG] [Cowork] [Dispatch] Permission prompts pinned to original device — mobile-originated Dispatch session continued on desktop still routes prompts back to mobile
- [BUG] Cowork VM fails to start - HCS 0x80070005, Credential Guard conflict
- [Bug Report] Unable to process request - no details provided
- [BUG] /loop dynamic mode: user questions interjected into ScheduleWakeup turns get zero assistant text — model only re-runs checks and re-schedules (claude-fable-5)
- [Bug] Content Safety Filter False Positive on ML Research Code
- [FEATURE] Show tool logs for nested sub-agents spawned within an orchestrator session
- [BUG] False positive content filter block when writing standard BSL 1.1 LICENSE file
- [Bug] Unable to determine issue - no bug report provided
- [Bug] Fable 5 model unavailable as advisor option
- [BUG] Claude code can no longer be installed or updated using native installer on IPV6-only networks
- [BUG] Claude Code v2.1.173: ui completely broken
- [BUG] Windows: cleanup `taskkill /PID x /T /F` spawned by Claude Code hangs forever and accumulates by the hundreds (process + memory leak)
- [FEATURE] /branch should register forked sessions as navigable entries in claude agents
- [BUG] Desktop Code tab (3P/Bedrock) never defers MCP tools — tool search works in standalone CLI but not in the embedded Code engine
- I don't have any bug report content to generate a GitHub issue title from. Please provide the bug report details so I can create an appropriate technical issue title.
- [Bug] Data Loss: Database content deleted after Claude Code update
- [FEATURE] Expose 1M-token context window in Cowork (currently 200K cap)
- [FEATURE] Native Cowork UI on iOS/Android — full client, not just Dispatch remote control
- Background tasks panel: completed tasks stay listed as Running and dismissals do not persist
- UX: disabling a project-scope plugin you're actively developing triggers an unexpected model-elevation safeguard
- [BUG] VSCode extension hangs indefinitely in devcontainer — CLAUDE_CODE_SSE_PORT stale after extension restart (regression in 2.1.170)
- [FEATURE] Automatic model routing based on task complexity — use Haiku/Sonnet/Opus dynamically instead of burning the session model on every turn
- [BUG] Scheduled routine silently skips execution — "next run" advances without firing
- [Bug] Pyright LSP plugin: pyright-langserver executable not found in PATH
- [BUG] Cowork — VM boots but guest never connects; HNS network silently not created (Windows 11 ARM64 / Snapdragon X)
- [FEATURE] disable spell check in claude windows app
- [Bug] Sonnet + Fable Advisor unavailable despite documented support
- [Bug] Command+C interrupt not working on macOS
- URL scheme to deep-link a notification click to a specific conversation
- Background subagents do not inherit Write/Edit permission allow rules from settings.json or settings.local.json
- [BUG] sandbox.autoAllowBashIfSandboxed: true from server-managed settings is not applied — all non-allowlisted commands prompt
- [BUG] Auto-compaction in a team-lead background session silently switched the main loop to teammateDefaultModel (durable until manual /model)
- [BUG] "Windows auto-updater EBUSY: running 245MB SEA locks its dir during swap"
- I appreciate the context, but I need the actual bug report or issue description to generate a GitHub issue title. Could you please provide: 1. What problem occurred while using Claude Code? 2. What error message(s) did you see? 3. What were you trying to
- [BUG] Claude Desktop (Windows): every MSIX auto-update leaks a container silo, app won't launch until reboot
- [Bug] Model switching fails on simple operations
- Fable 5 safety banner advises /feedback, but /feedback is unavailable in the VS Code extension (dead-end feedback path)
- [bug] [area:agents] [platform:macos] Agent(isolation: "worktree") post-creation setup silently rewrites...
- Headless -p run exits 0 mid-task: empty assistant turn after successful tool call treated as normal end-of-turn
- [FEATURE] Hide or make configurable the branch diff-stats badge in the Desktop status bar
- [BUG] [area:tool-use] [platform:macos] Write tool's full-file-replacement default causes irrecoverable data loss on governed, untracked state files — no append-only or protected-path mechanism available
- [Bug] Anthropic API Error: Overly Aggressive Content Filtering on Harmless Chess Terminology
- [BUG] Claude Desktop: uninstalled extension cannot be reinstalled (install hangs / no-op), state survives app reinstall
- Desktop app (macOS): image file attachments render as blank, non-interactive cards
- with chinese
- [BUG] Desktop (Windows 11): conversation silently duplicated into 15+ sessions; one typed message delivered to multiple sessions; parallel copies independently edited the same file
- v2.1.175: bottom input box / status line renders corrupted (stale duplicated lines, detached prompt, border drawn through typed text) — iTerm2, macOS
- You've hit your session limit · resets 9:20pm (Asia/Riyadh)
- [Bug] Safety filter triggering excessively on legitimate SWE tasks
- [Feature Request] Spend limit display misleading when set mid-cycle with pre-existing member accruals
- Windows 10: Stuck MSIX packages (0x80073CFA) block new installation (0x80073CFF) after interrupted auto-update
- [BUG] Windows: Active Claude Code agent pegs iGPU depending on visible Claude Desktop window size
- What is the intended purpose and server-side data handling of the /buddy companion's buddy_react endpoint?
- [BUG] Scheduled task creation fails on Windows with UNC-redirected Documents folder
- Ctrl+V paste still broken in legacy Windows console (PowerShell/conhost) - Jan 2026
- [BUG] Windows desktop app main process does ~95,000 file ops/sec → kernel non-paged pool leak (~14 GB/hr) → full system freeze. Regression in 1.11847.5
- Remote Control: Desktop SSH reconnect restarts the host daemon and kills all child sessions; Remote Control flag silently lost
- Remote Control: concurrent client attach forks the session (Epoch mismatch 409), orphaning the previously-attached client
- [BUG] StatusLine command receives open-but-never-written stdin pipe, blocking all stdin reads
- [Bug] Anthropic API Error: Model temporarily unavailable blocks auto mode safety checks
- [SYS-087] Fable 5 safety classifier false-positive on network/socket terminology forces mid-session model switch to Opus 4.8
- [BUG] Weekly usage limit banner appears persistently despite being at 31% usage
- You've hit your monthly spend limit · raise it at claude.ai/settings/usage
- [BUG] VSCode: ide_selection and ide_opened_file missing on first message when using a skill
- CI-monitor events fire from archived sessions and fan out to unrelated active sessions (stale session→PR binding)
- [BUG] opusplan/haiku in plan mode: subagents (e.g. Explore) inherit plan-upgrade model instead of their defined model
- Redirect messages stall behind long-running tool calls — /btw-style interrupt evaluation would fix this
- Feature request: model-driven interactive panel in the side panel (live dashboards, diagrams, pipelines)
- Wakeup-resumed turns intermittently serialize first tool call as literal <invoke> XML in text; with end_turn no retry is injected and ScheduleWakeup loops die silently
- [Feature Request] Add Projects feature and CLI parity to Claude Code desktop/mobile apps
- Compound-command detector flags read-only `cd && <cmd>` patterns, blocking common workflows
- VS Code extension: synchronous shell launch in resolveShellPath blocks event loop, causing extension host to be marked unresponsive
- [Bug Report] Unable to process empty bug report
- [BUG] Inconsistent usage-credit behavior between Claude Code terminal and web chat (Max 20x): CC accumulates cost but does not debit credits
- Bug: /btw command no longer works in VS Code
- Voice input: transcription rendered twice (duplicate text) in UI
- [Bug] Agent prioritizes emotional commentary over task execution
- [Bug] Claude Fable 5 model switching triggered by false positive safety classifier on legitimate security/privacy code
- [Bug] Model inverts instructions, modifying protected code sections instead of specified areas
- FIX THE DAMN AI
- [Feature Request] Add proactive context compaction before token limit with agent state preservation
- [Bug] Unable to process empty bug report
- Claude Desktop 1.12603.1 (Windows): installing a vendored .mcpb extension from the org directory hangs — "installDxtFromDirectory: reply was never sent"
- [BUG] The api.anthropic.com/authorize endpoint is broken and redirecting all Microsoft-365 MCP auth requests to a dead Google Drive page.
- [Feature Request] Optimize token consumption and response speed
- [BUG] Cowork: \outputs directory not created, causes immediate crash (Pro account)
- [FEATURE] Provide a prominent toggle to stop auto-submitting file contents
- Usage-limit gating is invisible to unattended sessions — agent-team orchestrator froze 5h with zero signal in transcript or terminal
- [BUG] [Accessibility] Permission dialogs fail WCAG 2.1.1: keyboard cannot reach non-default buttons in desktop app (focus escapes the dialog)
- [BUG] CVP-approved org: real-time cyber classifier false-positives at 0 tokens on authorized defensive-security work — Fable 5 blocked, Sonnet passes the identical task (request IDs included)
- [BUG] `/loop` scheduled cron fires at wrong time — offset from session start instead of declared cron minute
- Allow Claude to invoke /rename programmatically from within a session
- [BUG] Claude Code on the web stops streaming task progress in desktop WebKit browsers until refresh
- Background job sessions: assistant text written before AskUserQuestion is not shown to the user
- File viewer pane does not auto-refresh when file is edited by Claude
- [FEATURE] Cowork VM uses fixed HCS identifier, preventing concurrent workspaces across multiple Claude Desktop instances
- [BUG] darwin-x64 binary requires AVX2 since 2.1.113 — please ship bun-darwin-x64-baseline (like Linux/Windows already do)
- [Feature Request] Allow API access for legitimate cybersecurity and antivirus application development
- [BUG] macOS Dictation (double-Fn) stops working in the message input after a prompt is sent
- [FR] FleetView / task-list panel: show which model each agent is running
- [Feature Request] Support for security-focused code generation tasks
- [Bug Report] Unable to process issue - no report content provided
- [BUG] .mcpb file downloading but not getting extracted to Claude Extensions Folder
- [Bug] Umlauts replaced with ASCII characters in CLI responses
- [Bug] Anthropic API Error: Server rate limiting on requests
- [BUG] Enterprise allowedMcpServers policy completely blocks Claude Routines due to un-decoded session proxy URLs
- Cowork workspace fails with "VM connection timeout after 60 seconds" — CoworkScheduledTasks memory leak
- Allow MCP-provided Edit/Write tools to render diffs in the UI
- Feature request: Lazy/on-demand MCP loading triggered by skills
- Add setting to disable startup model announcement banners
- Feature Request: Support drag-and-drop file upload in VS Code extension
- [FEATURE] Allow /insights to include cloud sessions
- [Bug] Incorrect flagging logic causes unnecessary fallback to Opus model
- [FABLE]
- [Bug] Anthropic API Error: Safety measures blocking legitimate scientific research requests
- [BUG] Scheduled-task sessions surface inconsistently in Desktop sidebar: sometimes as unmarked ordinary chats, sometimes not at all; no badge, no running indicator, no reliable way to find them
- [Bug] Model unexpectedly downgraded to Opus 4.8 during security analysis request of internal app
- [BUG] 437 subagents is probably too many
- [Bug] Claude Code flags legitimate refactoring work as suspicious activity
- [BUG] VSCode extension ignores all managed settings (managed-settings.json, MDM profile, user settings.json) on macOS
- [Feature Request] Allow authorized security testing on user-owned infrastructure in Claude Code
- [Bug] Fable 5 safety measures incorrectly flag legitimate work instructions
- MCP HTTP OAuth: "client secret is invalid" with Google Desktop OAuth clients; SDK rejects valid credentials
- [Bug] API Requests Blocked by Safety Measures on Portfolio Use Case
- [Bug] Auto-update failed
- [BUG] Claude Desktop: extension install fully broken — "Install Extension" no-ops and drag-and-drop falls through the modal into the chat
- [Bug Report] Unable to process empty bug report submission
- [BUG] Claude Code, at least when using Opus 4.8, refuses to respond to anything, due to Usage Policy violation
- [BUG][bedrock] claude-opus-4-8[1m] from /model picker writes invalid model ID → 400 "The provided model identifier is invalid"
- [Bug] Fable model does not support security prompts or CTF/vulnerability testing
- [BUG] Desktop app: /mcp opens Connectors panel that omits project-scoped (.mcp.json) servers — no way to check their status
- [Bug] Model unexpectedly switched from Claude 3.5 Sonnet to Opus 4.8 during code review
- [Feature Request] Add chat archival functionality per project in agents screen [Feature Request] Improve slash command discoverability and categorization in CLI
- [BUG] `roots/list` is incorrect on Windows
- [Bug] Claude repetitively outputs same plan instead of simplifying with Fable x-high
- Terminal scrollback lost during long session — cannot scroll up to previous results (WSL2)
- [FEATURE] Surface tool-call request/response timestamps to the model so it can perceive elapsed time during long-running tools
- [Bug] Multi-line assistant text blocks intermittently not rendered in terminal output
- [BUG] # Bug report: Claude in Chrome via Cowork — all navigations instantly denied, no permission prompt ever renders
- Usage breakdown percentages are confusing — not clear they're independent
- [Bug] Claude Code hangs on billing prompt when monthly limit reached
- [Bug] Biology classifier false-positives on benign theory queries, incorrectly downgrading Fable to Opus
- [FEATURE] Antigravity Style Artifacts with Inline comments
- [BUG] Cowork (local-agent-mode) syncs and runs plugins disabled in enabledPlugins settings
- [BUG] Windows: screenshot clipboard paste fails ("No image found in clipboard") although the image is present in PNG+Bitmap formats
- [BUG] Desktop app injects computer-use MCP (27 tools + ~1.4K-token instruction block) into interactive sessions even when Computer Use is disabled
- Desktop: file viewer side panel never reloads a file the agent edits — stale content until tab is closed and reopened
- [BUG] Desktop preview pane: Reload button does not refresh stale page content after the file changes on disk (macOS)
- Auto-generated conversation titles ignore conversation language (always English)
- [BUG] .worktreeinclude copy drops/flattens symlinks — carried framework bundles and venvs arrive broken
- Opus 4.7 (Claude Code): scope creep + tool-state edge case during a refactor
- [Bug] Feedback UI appears above prompt blocking agent suggestions
- [BUG] Windows Desktop: main process permanently stops writing main.log once it hits the 10 MB rotation cap mid-run
- [BUG] Cowork sandbox fails to start on macOS after June 10 app update: "Workspace unavailable... Download failed"
- Fable 5: action-bias over answering — question-turns get tool actions and buried answers, user left 'dazzled'
- [BUG] Fable 5 confabulates external concurrent writes during filesystem MCP operations, session crashes and disappears
- [FEATURE] put MCP config under .claude/, like Kiro, Copilot, Cursor and the rest
- Model fabricates explanations for failures instead of investigating
- Model fabricates explanations for failures instead of diagnosing
- [FEATURE] "Pontificating" and "Exaggerating"
- Naming suggestion: give Claude Code the nickname "CiCi"
- Remote Control regression: server-injected `<system-reminder>Message sent at …>` prefix breaks slash commands from mobile/web
- [Feature Request] Rebrand Claude Code with user-friendly nickname for marketing
- [Bug] Safety filters incorrectly blocking legitimate code audit workflows
- [Bug] Anthropic API Error: Server rate limiting on requests
- Workflow tool: expose `effort` in agent() options (today you must create subagents just to pin effort)
- Agent registry is cached at session start; newly-created ~/.claude/agents files don't register until restart
- Overcautious refusal pattern for authorized kernel-driver QA steers users to worse tests
- [BUG] Claude Desktop will not launch on Macbook Air
- [DOCS] Align plugins-reference with sub-agents on recursive agents/ discovery
- [Feature Request] Add prompt analysis tools to understand user interaction patterns and intent
- [claude.ai] Settings → Usage page caused Chrome to hang with loop-like behavior (not reproducible after hard refresh — possibly stale cached bundle)
- [BUG] Claude Desktop ARM64 (MSIX): GPU process crashes on launch (exitCode 101457950), app exits — persists across GPU driver update and clean reinstall
- [Bug] Claude refusing legitimate medical/health content related to sexual wellness assessment
- Toggle between different accounts
- [BUG] Input line renders at wrong row (stale size model?) — any window resize instantly heals it; affects both renderers
- [Bug] Claude model auto-switches from selected model during security vulnerability analysis
- [Bug] Fable review mode incorrectly flags open source project analysis
- [FEATURE] Set default model for compaction (ie, I'm using Opus but /compact uses haiku)
- Not a bug — a thank-you note from a retired user in France
- Claude Desktop hard-codes SSH_AUTH_SOCK from "ssh -G ccd-identityagent-probe", breaking host-conditional IdentityAgent configs
- [FEATURE] /insights report: dark-mode support (prefers-color-scheme)
- `/clear` from the web remote control shows "Context cleared" but does not clear the session context
- [FEATURE] Idle / timeout / debounced hook event to enable customized handoffs, etc.
- [BUG] /context reports % against a 200k window on a 1M-context Opus model (false "over 100%" + spurious autocompact warning)
- Docs: clarify whether in-place forking (/branch, /btw → f) preserves or tears down the original session's in-flight background subagents/tasks
- [Feature Request] Add timeout warning and context preservation for execution_time_exceeded errors
- Feature request: programmatically focus/attach a session in the agents view (e.g. `claude agents --select <session-id>`)
- Background agent notifications route through wrong agent ID when launched sequentially (2.1.172+)
- [BUG] Truncated shell snapshot silently drops plugin bin/ PATH entries (partial snapshot used after builder is killed)
- [BUG] /feedback not working most of the time
- [Bug] Anthropic API Error: Fabble blocked /security-review endpoint
- [Feature Request] Allow tool usage filtering for AI agent safety research
- [BUG] Claude Desktop (Windows, MSIX) fails every launch — "untrusted mount point" (STATUS_REDIRECTION_NOT_TRUSTED); Squirrel/winget build works on same machine
- [FEATURE] Claude Code sessions cannot publish/update skills to the account Skills store - expose the Chat surface's "Save skill" flow
- AskUserQuestion race with in-flight typing: UI shows 'User declined to answer questions' but a selected answer is delivered to the model
- [Bug] Text rendering garbled in terminal on Ubuntu with AMD GPU
- [FEATURE] Expose context_management / clear_tool_uses_20250919 config for long-running agent sessions
- [Bug] Anthropic API Error: Safety filter triggers on legitimate security audit code review
- [Bug] Claude Code incorrectly downgrades to Opus for legitimate privacy/compliance scanning tool
- [BUG] The GitHub Actions integration on this repo closes threads as "stale" without warning
- I'm ready to help generate a GitHub issue title, but I don't see a bug report in your message. Could you please provide the actual bug report or issue description for Claude Code? Once you share the details of the problem, I'll create a concise, technica
- I'm ready to help generate a GitHub issue title for Claude Code. However, I don't see a specific bug report or issue description in your message. Could you please provide the actual bug report or issue details that you'd like me to create a title for? On
- Parallel subagent file edits clobber each other in the IDE diff editor (N−1 spurious permission-denials); diff-review lock is per-agent, not per-IDE-connection
- [Bug] Anthropic API Error: Server rate limiting on valid requests
- TUI silently drops assistant text emitted after tool results (only first short single-line mid-turn segment renders)
- [BUG] Desktop global "Auto-fix CI and address comments" toggle never applies to PRs created via gh from local sessions, and doesn't persist in claude_desktop_config.json
- False-positive dual-use safety flag on routine self-owned DevOps session (Fable 5, flagged 2x)
- [BUG] Session feedback prompt: "How is Claude doing this session?" answered with "Good" is grammatically mismatched
- [Feature Request] Support multi-language skill trigger phrases — non-English users cannot auto-trigger skills
- [FEATURE] Allow switching to dontAsk mode at runtime.
- VS Code extension: Effort slider disappears (instead of disabling) when usage limit forces model fallback
- [Bug] `/status` command Config output does not match documentation regarding configuration layers and settings display (re-opening)
- [Bug] Auto-downgrade from Fable to Opus triggered by false-positive safety flag on legitimate OSS repository
- [Bug] Safety classifier incorrectly triggers model downgrade to Opus for benign R&D discussion
- [BUG] Cowork: all tasks fail with retry loop then ECONNRESET — fails across networks incl. phone hotspot, survived clean reinstall, worked hours earlier same day (macOS 15.7.3, M3)
- [BUG] Workflow parallel() sub-agent loops forever emitting empty StructuredOutput calls; no retry cap and no per-agent timeout stalls the whole workflow
- [BUG] Desktop app telemetry missing user.email in of requests across all versions
- [Bug] Overly restrictive content filtering on local code review
- [Feature Request] Clarify API safety guidelines for privacy-focused local data handling
- [BUG] Auto-compaction fires without pre-compaction warning — silent context boundary causes summary-as-pending-action re-execution (regression from earlier versions)
Resolved issues (338)
- [Bug] Model incorrectly flags benign conversation and falls back to Opus
- [Bug] Computer use requests not sent through Remote Control approval interface
- [Bug] Unexpected High API Cost Usage
- Expose session usage-limit state to the model and hooks, with a configurable threshold and automatic post-pause resume
- [FEATURE] Runtime-enforced verification gates for Skills
- `claude plugin list --json` truncates stdout at 64 KB when piped and breaks VS Code plugin manager "available" list
- [BUG] Regression: VS Code "Manage Plugins" shows "No plugins available" with multiple marketplaces (64KB JSON truncation, previously fixed in 2.1.117 — refs #47237)
- [BUG] Subagent frontmatter `disallowedTools` silently ignores MCP glob patterns (`mcp__<server>__*`) while `tools` honors them
- CLAUDE_CODE_PLUGIN_KEEP_MARKETPLACE_ON_FAILURE still breaks fresh marketplace install on 2.1.169 (ref #43929)
- Data-quality: in-app /bug → LLM-titled pipeline yields ~1.5% content-empty issues naive metrics can't detect (measured, reproducible)
- [BUG] `/plugin` TUI shows `defaultEnabled: false` plugins as enabled (loader and CLI correctly report disabled)
- preview_start: autoPort always fails on Windows/Git Bash with "port in use" for all free ports
- preview_start breaks on Windows/Git Bash — autoPort returns "in use" for all free ports (regression)
- [BUG] Claude No longer works on New worktree
- [Bug] Anthropic API: Fable 5 safety classifier incorrectly flags all new sessions
- subscriptionType shows "pro" instead of "max" for Max 5x subscribers
- [BUG] IDE connection lost after /clear and /ide cannot reconnect (same as #55408, #28830)
- [Bug] Cybersecurity classifier false positives during authorized security audits of own codebase
- [FEATURE] check the generated code before to apply it
- [Feature Request] Option to disable automatic context compression
- [BUG]
- [BUG] Gamma connector OAuth completes but connector never registers as connected
- Claude randomly typed some chinese/japanese chars? "Let me do a final状态 check on the tree and tests."
- [BUG] Multi-window: "/" skill picker is mispositioned and Enter doesn't select (macOS)
- [MODEL]
- [BUG] /btw side chat opens as regular message instead of overlay
- Custom command statusLine never displayed on Windows (Git Bash) — Claude Code 2.1.156
- Cross-session message duplication: same user message (identical UUID+ms timestamp) delivered to multiple parallel sessions; duplicate JSONL log lines
- Opus 4.7 in Claude Code kicked paid EAS Build without authorization after explicit user rule
- Feature request: copy button for individual chat messages
- [Bug] /copy command corrupts UTF-8 text to latin-1 mojibake in clipboard write
- [Bug] /ultraplan fails on newly-created repos due to CLI repo-sync lag
- channels: mid-turn inbound injected into TUI input box instead of queuing as discrete turn
- Feature request: a producer-initiated "resume session" signal — a clean one-shot wake for background/external events
- Context lost after clicking software upgrade
- [FEATURE] Folders/Bins for normal chat
- VSCode extension /model picker shows Opus 4.6 as Default instead of Opus 4.8 1M for Claude Max account
- OTLP telemetry silently not emitted on v2.1.153 and v2.1.169 (macOS, interactive TUI)
- [BUG] Large tool outputs containing Japanese text become corrupted and expand into fabricated lines on Windows
- [BUG] Claude Code (VS Code ▎ extension) — copying a code block ▎ injects newlines into long commands
- [MODEL] opus 4.8
- [BUG] Cowork downloads Linux ELF binary on Intel macOS — exit code 132 (SIGILL)
- [Feature Request] AskUserQuestion option descriptions truncate on mobile, hiding key differentiators
- [FEATURE] "How is Claude doing this session" feels meaningless
- Recurring SIGILL/segfault in JSC GC (CodeBlockSet::clearCurrentlyExecutingAndRemoveDeadCodeBlocks) — bun 1.3.14
- effortLevel in settings.json not honored on session start
- Claude introduced JS syntax error in Python template, spent 2 sessions misdiagnosing
- [Bug Report: Unable to process request - insufficient information provided]
- Tab progress indicator stops permanently in one long-running session (Windows Terminal)
- [Bug] Anthropic API Error: Server rate limiting - temporary request limit
- [Bug] Anthropic API Error: Output blocked by content filtering policy
- [Bug Report] Unable to process empty issue submission
- [BUG]
- Feature request: Stop hook should receive assistant response text for output validation
- [Bug] Terminal display corruption after extended runtime
- You've hit your session limit · resets 3:40am (Asia/Karachi)
- Credit balance is too low
- [BUG] Desktop app: queued messages don't fire on their own — only flush when you send another message
- LSP tool not recognizing custom LSP server configuration (metals/Scala)
- [Feature] 允许用户配置每轮工具调用上限 (per-turn tool call limit)
- [BUG] exclamation mark corrupted in shell calls
- Recurring loop: 'an image in the conversation could not be processed and was removed' burns all tokens
- [Bug] Clipboard image paste fails with "no image found in the clipboard" error
- Assistant generated a fake "user" turn within its own response and then executed it as a user instruction on the next turn
- [BUG] Desktop app (2.1.168) remote Linux session: bridge connects but no ccd-cli session is ever spawned — hangs on "Starting session…"
- Skills injecting into system prompt even with plugin disabled — consuming 22% context at session start
- [BUG] Allow rules reject partial tool-name globs despite accepting full globs on the same server prefix
- [Bug] MCP Integration Error or Missing MCP Support
- [BUG] /resume lists no conversations when the project history dir is a symlink AND the repo has >1 git worktree
- I cannot generate a GitHub issue title because no error message or bug description was provided in English or French. Please share: 1. The actual error message you're encountering 2. What you were trying to do when the error occurred 3. Any relevant conte
- # Claude Code (Opus 4.7) — Diagnostic reasoning regression in long debugging sessions
- [BUG] Persistent "virtualization" banner in Claude Desktop Cowork when running under Parallels on Apple Silicon
- [Bug] Text output corruption in Kitty terminal with cell drift
- Third-party marketplace with autoUpdate:true not refreshed at session start; new release not picked up across restarts (v2.1.175)
- [Bug] Console billing: Tax calculation not resolving in checkout form
- [BUG billing] Got silently upgraded to Max plan from Pro without my consent. I got charged 200$ and no refund twice.
- [FEATURE] claude code api
- [BUG] Spell checker uses US English despite system language set to English (United Kingdom)
- [FEATURE] Usage Bar in Claude Desktop Header + Design Opportunity
- [BUG] [BUG] [Billing] $2,196 in runaway Auto-recharge charges on Max 5x plan — chat-only user, no Claude Code, no API keys — invoices missing from both claude.ai and Console billing
- [BUG] The model it's not working. Just crashed all my repos
- [BUG] VS Code OOM crash when Claude Code installed on SSH remote host
- [Bug] Skills loaded but not visible in Claude Code interface
- [Bug] Team memories feature not functioning correctly
- [Bug] Skills data lost or not persisting after session
- [Bug Report] Unable to generate issue title - no bug report content provided
- [BUG] Cowork "Download failed' on macOS Taho 26.4.1 M4
- [BUG] Cowork Linux sandbox fails to start — persistent for 24+ hours
- [BUG] claude-sonnet-4-6 regression: patch-on-patch reasoning on cascading state, requires explicit prompts to design properly
- [BUG] commit-commands:commit plugin loads full git diff into context — causes context bloat on large commits
- [FEATURE] Allow hooks to switch permission mode
- [BUG] Claude will return soon
- MCP collision check incorrectly suppresses claude.ai connectors that point to a different workspace than the same-named plugin — and there is no scriptable workaround
- Read/cat on config files leaks API keys into conversation transcript
- Wide-table fallback leaves stale bordered paint in scroll buffer alongside the key-value re-render
- Worktree isolation default-on causes confusing stale-deploy bug
- [BUG]Weekly Limits Are Draining Much Faster After the 5-Hour Limit Update
- MCP servers crash on macOS when accessing Bluetooth — missing NSBluetoothAlwaysUsageDescription in claude-code bundle
- [BUG] Claude Desktop - Co-Work not functioning
- [BUG] UI State Race Condition / App freeze during overlapping interactive requests
- Model recommended inline-secret command that leaks rescue admin password in Claude-isolation setup
- [BUG] Claude Code Memory leak after startup
- [Bug] Claude in Chrome - Login button flickers/loops repeatedly, unable to use Claude
- [BUG] Branched session re-renders prior slash command invocations as raw <command-message>/<command-name> XML tags
- [Bug] Statusline command not executed on Windows in v2.1.71
- Background session UI: 'describe a task for a new session' input stuck in conversation view with large blank space
- [BUG] Constantly hitting the "conversation is too long" error message with new chats despite only 16k/200k (8%) tokens used
- TUI leaks repeated fragments of current screen into terminal scrollback during a session, with rendering artifacts
- [BUG] First user message after `/clear` is rendered as a bare line inside the `/clear` turn's `<local-command-caveat>` framing — model treats the new message as stray output and ignores it
- feat: emit ANSI cursor shape escape sequences based on VIM mode
- Forked-skill Agent dispatch fails in VSCode extension 2.1.140 — sub-agent loses skill context and tool whitelist
- Agent tool: `isolation: worktree` switches parent worktree's branch instead of creating isolated worktree
- [FEATURE] Ignore-whitespace option for Edit tool diff display
- Code tab SSH remote session feature is broken.
- [BUG] VS CODE LIMITES RANGES VS CLAUDE CODE
- [FEATURE] Add latest model aliases for model setting in settings.json
- Surface-level fix tendency in long-running projects; previous-session directives not carrying into subsequent bug fixes
- [BUG] v2.1.140 VS Code extension on Windows: project skills missing from slash-command autocomplete (#9710 / #32331 cluster unfixed)
- [BUG] Severe native memory leak (~738 GB/h) — CLI fully unresponsive (slash commands frozen) on macOS
- Silent ANTHROPIC_API_KEY precedence shadows Max subscription auth
- [BUG] The agent automatically stops the task.
- [Bug] Claude 4.7 Opus degraded reasoning in issue triage tasks
- [BUG] computer-use request_access times out at 180s with no permission dialog
- [Bug] Cowork returns exit code 127 with "CoworkVM service not running" after update
- [BUG] VSCode extension: Default model picker shows Opus 4.6 but sets claude-sonnet-4-6
- [BUG] Claude Code hangs with Unable to connect to API (ECONNRESET) on every prompt — both CLI and in-app, persists across networks and clean reinstall
- `@`-mention synthetic Read injection doesn't satisfy Edit/Write 'must Read first' guard, forces redundant Read
- [BUG] spinnerVerbs setting not respected in VSCode extension
- [FEATURE] JetBrains plugin should support pasting/dragging images into the embedded terminal, parity with VS Code extension.
- [BUG] Desktop App Prompt Preview Overlay covers large portions of the output screen degrading UX and making it hard to read output.
- Welcome screen flashes and disappears on startup (v2.1.114, macOS)
- [Bug] Incorrect escape character in logger documentation example
- [FEATURE] Opusplan model with 1M tokens
- VSCode Extension: Interactive terminal integration (write to and read from VSCode terminal)
- [FEATURE] "Updated plan" should have the option to show the edit.
- [BUG] claude.ai/Design - no access on Max subscription, UK account.
- Claude Code API not connecting
- [BUG] Error: Claude Code process exited with code 3 when I tried to use the claude code plugin in visual studio code
- [BUG] Thinking blocks stripped from in-prompt context on every turn in active Opus 4.7 sessions (no idle-resume trigger)
- [Bug] Extreme token consumption in API requests
- [FEATURE] Sensitive data redaction in transcripts shared for debugging
- [Desktop App] Allow customizing Enter key behavior (newline vs submit)
- [DOCS] 5 hour limit removed, but weekly limit remains? No more capacity just able to hit limit faster?
- `claude mcp remove` expands `${VAR}` env-var references inline in `.mcp.json`, leaking secrets
- [BUG] VS Code: abandoned chat panel creates stub session, "No conversation found" fires on every reload
- [FEATURE] /insights: Feature Fluency report card — per-feature adoption grades and gap analysis
- [FEATURE] Per-skill lazy/on-demand loading for skills (slash commands)
- [BUG] LocalMcpServerManager kills active MCP process on session focus change — in-flight tool calls lost with 'Tool result could not be submitted'
- mcpServers config gets corrupted with nested "mcpServers" key in ~/.claude.json
- [FEATURE] Ship a built-in Terse output style and tighten Default style's commenting behavior
- [BUG] Failed to authenticate. API Error: 401 {"type":"error","error":{"type":"authentication_error","message":"Invalid authentication credentials"},"request_id":"xxxxxxxxx"}
- [BUG] 401 Auth Loop - All commands fail including /login, Windows, Pro subscription**
- Status line shows empty `current: · latest:` with no version values
- Plugin install/update writes ~/.claude/plugins/npm-cache/package.json without truncating, leaving trailing bytes from previous content
- Bash tool omits `< /dev/null` when user command has any `<`/`<<` redirect, leaking parent stdin to children
- [FEATURE] Claude Code Web should not impersonate users on GitHub
- Agent view: surface PR CI status on PR cards (✅ / 🔄 / ❌)
- [FEATURE] Vim/emacs-style row navigation in agent view (j/k, ctrl+n/p)
- [FEATURE] Send messages to running agents without stopping them
- [FEATURE] Agent view should use available terminal height instead of truncating behind 'show more'
- Misleading keyboard shortcut labels on macOS: 'Alt' should be 'Option' (⌥)
- Agent tool subagent does not load working directory's CLAUDE.md — inherits main session context only
- Base64 data URI in tool output triggers 'Could not process image' API error, poisons conversation
- [DOCS] Clarify recursive subdirectory discovery in ~/.claude/agents/ and .claude/agents/
- [BUG] Claude Code in Claude Desktop fails to work with local projects (macos)
- Ctrl-G external editor spawn regression introduced between 2.1.132 and 2.1.138 (Windows + Cygwin/MSYS2)
- Subagent `model:` frontmatter pin silently ignored in 2.1.117
- [Perf/Cost] Tool catalog (~24k tokens) sent uncached on every API request — cache_control breakpoint missing
- [FEATURE] Agents View: show branch name, merge status, and mergeable state per session row
- AgentView: keep-alive / external-event-trigger hook for long-running MCP-driven sessions
- [FEATURE] View-level meta-agent in the Agents View — conversational oversight across running sessions
- [BUG]Claude Code starts but is unresponsive
- Feature Request: Show session status (running/pending/completed) in session list
- [FEATURE] Compose multiple Skills/Commands in a single user message
- [BUG] Failed to start Claude's workspace — Request error: net::ERR_CONNECTION_CLOSED
- [FEATURE] Support Agent/Prompt Hooks Triggered Immediately After Claude Prompts for User Input
- [MODEL] Sonnet 4.6 DOES NOT FOLLOW GITIGNORE AND USES -F BY DEFAULT
- [BUG] Desktop App Crashing and Not working At all
- Feature request: PreToolUse hooks need a silent-deny permission decision for non-interactive / dontAsk sessions
- [BUG] M-d / Alt+D (delete-word-forward) consumes trailing whitespace, diverging from expected behavior for readline / MacOS / Emacs
- [Bug] UI freezes and blurs when selecting GitHub repository items
- Feedback: /ultraplan friction — cwd locked at session start, web sandbox clones only once
- settings.local.json: 'Yes, allow' permission prompt writes malformed Read(//abs/path) with double leading slash
- [Bug] Auto-added Read/Edit permissions generate double-slash paths (//Users/... instead of /Users/...)
- Feedback: /ultraplan is positioned as deeper planning but is really a fresh-container Claude Code
- [BUG] fail to create custom style in a team plan
- [BUG] [Desktop App Code tab / macOS] Prompt preview overlay covers diff bar and permission prompts (Desktop App v1.7196.0)
- [BUG] Claude Desktop fails to launch on Windows 10 — EBADF renameSync error on git-worktrees.json
- [BUG] v2.1.141 regression: enterprise managed-mcp.json blocks Claude-in-Chrome dynamic MCP setup (worked in 2.1.140)
- Claude app causes macOS to terminate all sessions due to excessive VM disk writes (34 GB in one day)
- Extension host crash: Webview is disposed thrown from setTimeout in extension.js (v2.1.141)
- opus Skill rewrites: ignored own /verify skill, made unverified claims, regressed to prose summaries
- bug : Remote Control: /rc loops "connecting… → disconnected", never reaches Connected (Windows 11, Pro)
- Middle-click paste fires twice (on mousedown and mouseup) in VS Code integrated terminal — regression in 2.1.175
- [Bug] Fable 5 model unavailable as advisor option
- [BUG] VSCode extension hangs indefinitely in devcontainer — CLAUDE_CODE_SSE_PORT stale after extension restart (regression in 2.1.170)
- [BUG] Scheduled routine silently skips execution — "next run" advances without firing
- [BUG] Claude Desktop (Windows): every MSIX auto-update leaks a container silo, app won't launch until reboot
- [BUG] Cowork — Global user instructions edits don't persist across sessions; per-session CLAUDE.md caches retain stale content
- Redirect messages stall behind long-running tool calls — /btw-style interrupt evaluation would fix this
- [BUG] Failed to resume session: EINVAL: invalid argument, stat 'C:\Users\username
- Claude Desktop 1.12603.1 (Windows): installing a vendored .mcpb extension from the org directory hangs — "installDxtFromDirectory: reply was never sent"
- [BUG] darwin-x64 binary requires AVX2 since 2.1.113 — please ship bun-darwin-x64-baseline (like Linux/Windows already do)
- [BUG] .mcpb file downloading but not getting extracted to Claude Extensions Folder
- [BUG] `roots/list` is incorrect on Windows
- [BUG] claude code TUI has gone down hill
- [BUG] Scheduled task creation fails on Windows with UNC-redirected Documents folder
- [BUG] Channel-notification turns (--dangerously-load-development-channels) not written to session JSONL in 2.1.172/2.1.173
- [BUG] /context reports % against a 200k window on a 1M-context Opus model (false "over 100%" + spurious autocompact warning)
- [BUG] Atlassian connector OAuth callback fails - "연결할 수 없음"
- [BUG] Claude Desktop macOS window stays always on top of other apps
- [BUG] Claude desktop Windows ARM Version 1.11187.4 (584005) Broke PDF Reading
- [BUG] Voice mode unable to access project memory, instructions, and attachments on Claude Web
- [BUG] Claude Desktop (macOS) ignores MCP tools/list pagination — tools past page 1 never load
- [FEATURE] Claude Desktop Transcript View settings to expand Code changes only
- [Bug] Claude generating hallucinated fixes for non-existent issues
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [Bug Report: Unable to determine issue from empty submission]
- [Bug] Fable 5 incorrectly routes legitimate genomics research queries to Opus 4.8
- [BUG] Typst (.typ) files show "No content available" in Live UI preview panel
- [FEATURE] OAuth tokens should be revoked server-side on claude logout and on devcontainer shutdown to support secure workflows
- [BUG] User skill not appearing in Claude Desktop UI despite correct file placement (Windows, post-April 23 regression)
- [BUG] Workspace trust dialog asks on every session startup, doesn't persist acceptance
- [BUG]Token compensation request — Auto Mode classifier outage
- [Bug] Organization marketplace not appearing in plugin list after admin console deployment
- [BUG] 二进制加密问题
- [FEATURE] Add --no-hooks flag to claude -p (non-interactive mode without breaking OAuth/subscription auth)
- /branch from a rewound session leaks rewinded messages into the new session's /rewind picker
- [BUG] Windows installer wants admin rights, can't install as standard user
- [Bug] /mcp dialog no longer shows claude.ai MCP marketplace list (v2.1.140)
- [BUG] Bun panic: invalid enum value on Windows win11_dt + WSL reg.exe ENOEXEC error
- [BUG] Hard crash at startup on WSL2 when binfmt_misc has no WSLInterop entry (ENOEXEC spawning reg.exe)
- [BUG] [Windows] Claude Code 2.1.139 still hangs after all documented workarounds (spawn EINVAL, blank -p output, interactive hang)
- Account-level connectors no longer appear in `claude mcp list` (Vertex AI auth)
- [MODEL] Model dismisses valid contextual questions during active development session
- [BUG] claude -p silently exits with no output on Windows 11 (v2.1.131)
- [BUG] Usage Reset Not Triggering at Scheduled Time
- Chrome extension not connecting to Claude Code despite correct setup
- [BUG] Claude Desktop MSIX fails to launch — EBADF on renameSync(git-worktrees.json) — on Windows 11 25H2 Build 26200
- Opus 4.7 regression: weaker keyboard-language typo detection + less precise memory/MD file writes vs Opus 4.6
- [BUG] MCP tool calls with Optional list parameters fail — list serialized as string
- [BUG] VSCode extension: auto-scroll during streaming prevents scrolling up
- Microsoft 365 MCP connector: redirect URI mismatch (AADSTS50011)
- [BUG] BUG: Cowork Linux sandbox fails to start on Windows — persists after all Hyper-V service fixes
- [BUG] model anchors on dangerouslyDisableSandbox: true and keeps applying it to unrelated read-only commands
- [FEATURE] Desktop App sessions should either sync to cloud or clearly indicate they are local-only.
- Clipboard-pasted screenshots are treated as duplicates in VS Code extension
- [BUG] USAGE LIMIT UNUSUAL BEHAVIOUR
- [Feature] OnApiRetry hook event for observability and control over internal retry behavior
- [BUG] Google Drive MCP connector — create_file with explicit parentId fails "User cannot add children to specified folder" — regression after May 7 incident
- [Bug] Claude generates fabricated content instead of requesting clarification
- [Bug] Agent suggests destructive commands without safety warnings
- Opus 4.7 over-hedges and defers action on feature-scoping conversations, ignoring user-supplied "be proactive" instructions
- [BUG] Claude Chrome extension renders black screen with no tool access in Vivaldi browser (Linux)
- TodoWrite system-reminder fires too aggressively when an active todo list exists
- [BUG] App intermittently goes permanently blank/white during active session — no crash logged, no recovery
- "Waiting for input" notification fires on Windows when Claude is not waiting for input
- [BUG] Claude Chrome Extension - unable to Authorize
- [Bug] Weekly all-models cap metering cache_read_input_tokens at full weight vs. 5h bucket discount
- [FEATURE] Expose context_management / clear_tool_uses_20250919 config for long-running agent sessions
- [BUG] Shell: bash (use Unix shell syntax, not Windows) 这个程序内置的提示存在歧义会导致理解成错误的含义
- [BUG] Memory leak in short, lightweight sessions (v2.1.104, macOS ARM64, 32GB)
- [BUG] disabledMcpjsonServers disabled when listing incorrect mcp value
- [BUG] Claude desktop loses most mcp connections
- Project list shows deleted projects that cannot be removed
- [BUG] /release-notes changelog cache not refreshing — stuck at 2.1.79 since March 2026
- [Bug] /ultrareview crashes on large diffs without producing findings or billable output
- [Feature Request] Claude should know which client/interface it's running in
- [Bug] Agent Teams auto-approve plans in CLI mode without user confirmation
- /clear orphans background tasks by modifying session state in place
- [BUG] Claude Code Extension open new editor window (won't stay in secondary side bar)
- [BUG] Webview: Monaco editor worker loading fails, triggers "Unhandled case" banner and interrupts session
- [BUG] `claude -p` / `claude --print` changes Claude Code product identity to Agent SDK identity
- [BUG] `claude -p` / `claude --print` exposes a different tool/default-prompt contract than TUI
- [DOCS] Document `claude -p` / `claude --print` parity boundaries and route known print/headless issues
- [FEATURE]
- SessionStart hooks defined in settings.json silently not invoked on Windows (v2.1.141)
- [BUG] Tips for getting started box shows truncated text even at full screen
- [BUG] Skill invocation bypassed on explicit user command despite CLAUDE.md mandate
- ultrareview crashed before producing findings
- [Bug] Thinking blocks rendered as conversation messages in output
- Auto-mode safety classifier denies the Co-Authored-By trailer that the Bash tool's own system prompt instructs the model to add
- [Bug] Unable to add or select project folder in Windows Desktop client v1.7196.0
- [FEATURE] Accept edit with details
- [BUG] Sends effort parameter when using custom model
- Feature request: memory decay and freshness — salience scoring or TTL for auto-memory files
- Desktop app persists session metadata before any user message → 'No messages yet' zombies clutter Recents
- Add setting to disable IDE "user opened file" context injection
- [Bug] Agent fails to search the internet, doesn't look at git history for chesterton's fence, and misses vulnerable package version off-by-one errors in code vulnerability
- Spacing bug: ⏸ plan mode indicator renders without enough space after the symbol
- [FEATURE] Cryptographically signed harness blocks to distinguish Claude Code injections from user content
- [BUG] REPL-level context limit check uses stale post-compact token accounting, blocks input despite 73% free context (1M window)
- autoMemoryDirectory: default global memory dir still created each session
- [Bug] DuckDB-WASM Parquet Query Optimization Not Utilized
- [FEATURE] CLAUDE_CODE_PERMISSION_MODE env var
- SessionStart hook only fires for the first ~5 project switches per Claude Code Desktop launch
- Auto-compaction fires without pre-compaction warning (regression from earlier versions)
- PostToolUse:Edit IDE diagnostics: 'code' field returns literal '[object Object]' instead of the diagnostic code string
- PreToolUse hooks should still be able to fire in bypass / dangerously-skip-permissions mode
- Remote control: toggle_remote_control completes silently — no local banner, no handshake confirmation, but phone-side registers connection (split-brain)
- VS Code extension: toggle_remote_control sent from webview, no confirmation event logged, banner-above-textbox never appears (yet phone registers session)
- Enhancement: surface local resource pressure (RAM/CPU/disk) below the input bar
- [BUG] While in planning mode model command tricks agent into thinking it's running Haiku but it is not
- [BUG] Auto-mode classifier: session-level trust decay + hard-rail misattributed as user rejection (Windows, v2.1.138)
- Refund request: repeated memory-rule violations in single Opus 4.7 session
- /usage command lost its progress bar — now opens settings dialog instead
- /doctor false positive: Plugin validation uses incorrect marketplace identifier format
- [BUG]
- [FEATURE] CLAUDE_CODE_OUTPUT_STYLE env var
- Ukrainian generation produces Russianisms / calques even with explicit 'avoid Russianisms' instruction — offering verified corpus + eval set
- [FEATURE] Feature request: font size control in VSCode extension panel
- [BUG] Unable to install Claude Cowork plugin on PC's. Affecting all PC user
- [BUG] [BUG] SSH-Remote ccd-cli 2.1.128 still exits with code 1 on initialize with hookCallbackIds — regression of #50698
- [BUG] can't connect claude agent to Xcode intelligence, but can connect claude chat to Xcode
- VSCode/Cursor extension: permission-IPC race kills sessions, error rendered as "[object Object]"
- Claude desktop 1.7196.0 (macOS) bounces in dock, no window, exits silently after upgrade
- [BUG] Windows: Cowork browser automation tools missing after Claude Desktop update — computer-use tools never register
- [Bug] Request timeout after extended execution duration
- [BUG] Context Compaction Generates Fabricated Pending Tasks Leading to Unintended Action Initiation
- [BUG] Renderer shows "Unhandled case: [object Object]" instead of context-limit error message when API rejects an oversized request
- [BUG] Cowork: New project with name matching selected context folder silently creates empty nested subfolder and mounts it instead of the context folder
- [BUG] mcp auth not automatically launching browser on Windows
- [BUG] /ultrareview crashes with "Review crashed unexpectedly" — 2 credits consumed with no output
- [BUG] Claude Code Web — All sessions destroyed during April 28 outage, account-level server-side corruption
- [BUG] Claude Code — Multiple critical errors causing extreme user frustration and stress
- Enhancement: expose CLAUDE_SESSION_ID environment variable in hooks and tool subprocesses
- Permission prompt auto-accepts when pressing Delete/Backspace while editing the "additional details" input
- [BUG] Claude Code Desktop crashes with "process exited with code 1" - binary works fine in terminal
- [Bug] `/status` command Config output does not match documentation regarding configuration layers and settings display
- Tasks sidebar: completed background bash tasks stay as "Running"; dismiss (X) button is unresponsive
- Auto-scroll behavior interrupts reading in chat window
- [BUG] Cowork rises CPU temperature to 83°C even when idle
- Webview crash "Unhandled case: [object Object]" — markdownlint diagnostics with non-string `code` field
- [DOCS] Document that `DEBUG` environment variable enables debug mode
- [Feature Request] Configurable permission mode cycle for shift+tab
- I'm ready to help generate a GitHub issue title for Claude Code. However, I don't see a specific bug report or issue description in your message. Could you please provide the actual bug report or issue details that you'd like me to create a title for? On
- [BUG] Slash command autocomplete stops filtering after a '.' (period) in skill/command name
- Slash-command menu returns zero matches when query contains `.` (regression 2.1.160 → 2.1.161)