Digest / May 31, 2026
Claude Code Issues Digest — May 31, 2026
264 new issues filed ·
422 resolved in anthropics/claude-code (UTC day).
New issues (264)
- [BUG] 5h token usage massivly outstripping actual context (issue started about 8:45pm BST)
- [Bug] Anthropic API Error: Cannot modify thinking blocks in assistant messages
- Tool-result envelope injected into tool-call input channel during parallel tool batches (corruption + cascade cancellation)
- [BUG] Auto-compact not triggering before 1M context limit error in Cowork mode
- [Bug] Claude Opus 4.8 emits malformed tool calls with parsing errors
- Add setting to disable or override animated spinner text
- [Bug] Tool-call parser rejects indented XML tags with vague "malformed" error
- Claude omits modified files when staging commits with new files
- [Bug] Anthropic API Error: Server Rate Limiting on Recurring Requests
- [Bug] Frequent API disconnects and excessive token consumption with Claude Opus 4.8
- [Bug] Model generating incorrect bash commands and false success reports
- # [BUG] [Windows] Auto-updater stuck in reinstall loop: daemon self-restarts on mtime change and reinstalls the same version repeatedly, flashing cmd windows every few minutes
- Allow running shell commands during active turn without canceling
- Tool calls leak as raw text ('call' + args) instead of executing — frequent on Opus 4.8
- I don't see a bug report or feature request in your message. This appears to be a question about Claude Code's behavior rather than an issue report. If you're asking about how Claude Code prioritizes different types of instructions (rules vs claude.md fil
- Tool calls intermittently emitted as literal text (stray 'court' + raw <invoke>) instead of executing — mostly Edit/Read (Opus, CLI)
- [BUG] Old sessions do not open, also cannot start new sessions
- Opus (CLI) took down a node's DNS by deploying a k8s LoadBalancer on :53, after dismissing on-screen evidence the port was the host resolver
- Session titles generated in English despite locale set to ko-KR
- [BUG] Repeated malformed tool-call markup (stray count before <invoke>) within one session — commands silently dropped
- Headless `claude -p` returns 400 when a connector tool has a top-level anyOf/oneOf/allOf in input_schema
- Better usage model (pay to push to reset)
- Shift+Enter to Control+J in Terminal
- scheduled-task sessions: intermittent silent boot failures + mid-session bash-bridge degradation
- [BUG] API Error: Usage credits required for 1M context
- [Bug] Claude hallucinating code without repository context validation
- Hook stdout shown + suppressOutput:true ignored in 2.1.158 (regressed from 2.1.157)
- Hook stdout shown + suppressOutput:true ignored in 2.1.158 (regressed from 2.1.157)
- AskUserQuestion overlay should be minimizable / dockable — it blocks the message it refers to
- [Bug] advisor() silently returns no result (red circle, ~4s, no advisor_tool_result) when transcript exceeds the 100-image API limit
- [Bug] Claude Code argues unnecessarily about spec-first workflow requirements
- [BUG] terminal-notifier 桌面通知在权限弹窗时不触发
- [BUG] Security-guard generation defaults to allow-list, ships incomplete fixes, and relocates vulnerabilities; only converges under looped adversarial review.
- [Bug] Claude Code unresponsive or not generating responses
- Spurious "tool call was malformed and could not be parsed" appended after successful tool calls
- Auto-mode classifier: `allow` rules silently ignored for default-branch protection; self-modification block contradicts documented escape hatch
- [Bug] Opus 4.8 responses not displayed after tool use — quota consumed with no output
- GitHub integration connects via OAuth only — private repos inaccessible in remote agents
- [Bug] Workflow stuck in idle state for 9-10 minutes
- [Bug] Claude Code advisor process crashing or becoming unresponsive
- Native CLI 2.1.152 consumes 8 GB+ RAM on regex with bounded repetition over minified single-line input
- [BUG] Remote Control (Android version) always fails to deliver .APK files
- [Bug] Tool output replay corruption with batched parallel calls causes phantom state
- v2.1.158 (Opus 4.8): main session missing Grep/Glob, Bash storms, duplicate Bash commands, fabricated branches/commits/files, excessive shell text commands
- [FEATURE] honesty exemption
- [BUG] Built-in local slash command output (e.g. /context) doesn't create a rewind checkpoint, while plugin commands do
- [Feature Request] Add window title indicator for Claude Code prompt state (green/red/yellow LED)
- Tool results arrive delayed/invisibly — model appears frozen during long Bash/SSH calls (Opus 4.8, VSCode extension)
- [FEATURE] Add a True Isolated / Incognito Chat Mode with User-Controlled Privacy Context Switch
- IntelliJ plugin: change chat panel color when not in focus
- [BUG] Session limits maxing out on their own, without any interaction from user.
- In-flight /btw side conversation silently lost on Code -> Cowork -> Code switch (data loss)
- Skill tool invocation triggers bypass_permissions_gate_blocked, downgrading session from bypassPermissions to acceptEdits
- [BUG]
- [Bug] Anthropic API Error 400: System message in messages array breaks session unrecoverably
- [BUG] Invalid Request
- [Feature Request] Improve error recovery and self-correction during multi-step operations
- Opus 4.8 (1M): tool-call emission repeatedly malformed (stray 'court' token + dropped namespace), renders as text and does not execute, no self-correction across ~10 retries
- Bundled GrowthBook clientKey intermittently returns 400 "Invalid API Key" → all feature flags fail → channels & Remote Control unavailable
- Claude Opus over-engineers simple tasks in agentic/CLI mode, wasting tokens
- [BUG] Opus 4.8 medium effort spends 46k output tokens on hidden thinking for a simple coding turn
- [FEATURE] I really wish I could minimize the questions popup window
- [BUG] .claude/CLAUDE.local.md not being loaded
- No warning when selected model is incompatible with xhigh/Ultracode effort level
- [BUG]
- [BUG] Calling advisor tool crash or stop the claude code with following API error Unsupported content type: server_tool_use
- [Bug] Text formatting lost when returning from /plugins command
- Claude repeatedly ignores tool rejections and burns user tokens
- Claude Code (Opus 4.7) silently narrows user-assigned task scope using self-generated safety heuristic, even when explicit in-session directive and persistent memory forbid it; laundered through external reviewer agent
- [BUG] Plan-mode read-only guarantee is bypassed by spawned subagents — auto mode classifier governs subagents instead of plan mode.
- [BUG] Visible console windows pop up for every bash command on Windows (Git Bash) — missing windowsHide flag
- [Bug] Model hallucinates prompt injection narrative to rationalize own tool call denial
- Feature request: Bash tool should respect user's configured shell
- [Bug] Claude ignores CLAUDE.md language rule and responds in Ukrainian despite English input
- [BUG] Invalid Oath - Claude Code
- [BUG] VS Code extension subprocess init timeout on WSL2/Ubuntu 26.04 — CLI fully healthy (stream-json handshake works from terminal)
- [BUG] Claude does not let me change
- [FEATURE] Add PermissionResolved hook event
- Feedback: noticeable reliability regression — agent edits from memory, silent Edit failures shipped broken code to prod
- [DOCS] Wrong and deprecated npm install command shown on claude.ai/code
- Channel splash shows 'plugin not installed' for all plugin-type channels despite channels working correctly (v2.1.145)
- Claude Code assistant repeatedly takes unauthorized actions, ignores stored behavioral rules
- Claude Code wastes user quota with repeated trial-and-error instead of verifying calculations before coding
- Regression in 2.1.150-2.1.158: malformed tool_use blocks not parseable, retry-resistant
- Workflow counts API-errored subagents as "completed"; 429/500/529 storm under multi-agent load
- [Feature Request] Require confirmation for open questions before presenting plan start options
- [Bug] Tool-call markup leaks into assistant text instead of tool_use block causing silent no-op
- [FEATURE] PostAssistantResponse
- Feature request: verbatim/bash-only mode for slash commands
- [BUG] Linux sandbox: proxy UNIX sockets not visible to subprocesses (missing bind after --tmpfs /tmp)
- [FEATURE] unarchive the archived sessions in Claude Code.. i accidentally archived my sessions
- [BUG] everything broke down on windos desktop app
- [FEATURE] HTML viewer and .md edit in Claude Desktop App
- [BUG] You ran our of memory
- [Bug] Tool result hallucination in long agentic sessions with batched tool calls
- [Bug] Opus 4.8: parallel Bash siblings stuck on Waiting… and render out of order behind a long-running call
- [BUG] - ultrareview failed and consumed free use credit
- After /compact on Opus 4.8 (1M), whole tool calls emitted as <invoke> text instead of executing (not seen on 4.7)
- `/desktop` fails: "Cannot determine working directory … transcript may be incomplete" after EnterWorktree→ExitWorktree (transcript fragmented across project dirs; bridge-session record has no cwd)
- TaskCreate 'task tools haven't been used recently' system-reminder fires repeatedly per long session — needs suppression knob (cross-repo from AaaS-Love/.claude#2048)
- [Feature/UX] Yellow notification dot never clears on read + no way to configure dot behaviour
- Token waste bug: Workflow spawned 44 parallel agents to read files instead of using git clone (2M tokens vs ~1000)
- [Feature Request] Restore deprecated Claude Opus versions (4.6, 4.7) with deprecation warnings
- Claude Pro: 1M context limit breaks active workflows without warning
- [BUG] macOS: npm/Homebrew install places native binary at bin/claude.exe, causing recurring "claude.exe would like to access data from other apps" TCC prompts
- [MODEL] False-positive cyber-safety block on benign Minecraft mod documentation task
- Unsent draft text lost when switching sessions (VS Code extension)
- I'm ready to help generate a GitHub issue title for Claude Code, but I don't see a bug report in your message. Please provide the bug report details, and I'll create a concise technical issue title following the guidelines you've outlined.
- [BUG] Title: Auth URL line-wraps in JetBrains IDE terminal, making clickable link invalid
- Regression in 2.1.158: claude -p hangs waiting for stdin EOF on Termux/arm64
- [Bug Report: Model performance degradation on project tasks
- [Feature Request] Message queuing for sequential task execution in VSCodeClaude extension
- [Feature Request] Background task execution without conversation blocking in VSCode Claude extension
- Bug: automatic cleanup deleted 6 active sessions without warning — token loss
- [FEATURE] Terminal (TUI): inline colored annotations + deferred batch reply to specific blocks of Claude's output (reviving #45904)
- [BUG] Image paste (Alt+V) always fails under WSL when login shell is fish — clipboard command run via $SHELL, which rejects "$(...)" in command position
- [Bug] Tool-layer unreliability in long/compacted sessions: duplicated results, delayed delivery, empty/garbled file reads
- V8 out-of-memory crashes (SIGABRT) freeze/kill the TUI during long sessions on macOS
- Session cut mid-task by credit limit — no warning, context lost, costly model switch
- [MODEL] Opus 4.8 (xhigh): claims work verified when it isn't, self-contradicts, and over-parallelizes contradictory tool calls — on a low-complexity codebase
- [BUG] 400 "thinking/redacted_thinking blocks cannot be modified" bricks session after large multi-tool turns
- [BUG] Agent View (Linux): mouse wheel sends arrow keys instead of scrolling attached session — works on macOS, fails on all Linux terminals
- iPad browser (vscode.dev): newlines not rendered in message display
- [Bug] Increased tool call failures and retry attempts with Opus model
- [Bug] Tool use regression in version 4.8 causing premature termination and parallel execution issues
- [Bug] Bash tool executes commands multiple times with duplicate outputs
- [Bug] Tool results delivered with multi-turn lag and flushed in batch
- [BUG] /init silently misses source files in repos with many files due to Glob truncation
- [Bug] Reasoning interruption in ultramode v4.8+ requires manual restart
- [BUG] Android app: web-search permission modal fails to appear in remote-control Code session, leaving the model stuck in an execution loop (recoverable from PC, not from Android)
- UserPromptSubmit hook failure blocks all input when plugin cache file is missing [Errno 2]
- Feature: first-class GitCommit action (bypass Bash argv for commit messages)
- [BUG]
- [BUG]
- Claude Code repeatedly made unauthorized destructive changes, ignored explicit rules, and permanently destroyed user data across multiple sessions[MODEL]
- Tool results delayed and duplicated in interactive sessions (v2.1.158, Linux, multiple concurrent sessions)
- Server-side tool (server_tool_use) emitted but no result returned; response reports stop_reason: end_turn
- False positive safety block on legitimate security audit tasks (Supabase RLS review, pnpm audit)
- [FEATURE] programmatic conversation renaming in Claude Cowork
- [Bug] Single-use OAuth refresh token consumed by test copy invalidates production token
- [BUG] Windows: stopShellPty on tab switch kills session with exit code 4294967295 — regression in May 27+ build
- [BUG] Cowork "Download failed" on macOS — claude-code-vm runtime missing from app bundle
- Regression (since 2026-05-29): intermittent "tool call was malformed and could not be parsed" — tool_use block absent on a stop_reason=tool_use turn
- [MODEL] Claude Code + Opus 4.8 frequently echos hello world during planning
- Expose a way to disable parallel tool execution (the API's disable_parallel_tool_use) in Claude Code
- [FEATURE] Decouple LSP diagnostic auto-push from the LSP tool (keep navigation, drop publishDiagnostics)
- [BUG] typescript-lsp pushes stale diagnostics on 2.1.158 (post-#17979) in TS project-references/composite workspaces
- [BUG] Remote-control sessions (browser/mobile) silently drop attachments — shown as sent, never reach the model
- run_in_background silently kills active processes and reports success (exit code 0)
- iOS Dispatch sessions from mobile fail to spawn reliably
- [BUG] Unable to copy `%PDF-1.6` text
- Dispatch needs user-level persistent persona / identity injection
- [FEATURE] Multiple account
- Parallel tool-call batch cascades "Cancelled" to independent calls when one fails
- [Bug] Parallel tool calls cancel all siblings on single error, which appears to cause model confusion
- [BUG] naming convention should not use "-" as the replacement for "/" as it creates file and directories that start with "-".
- [Bug] Agent stalls without token consumption during execution
- Advisor tool completes in UI but result never injected into model context (under Remote Control)
- [BUG] Claude Code query worker dies silently on Windows — `--version` works, but every query produces no output (exit 0); VS Code extension shows "Query closed before response received"
- macOS: keybinding hint shows 'meta+w' instead of 'Option+w' (⌥W)
- [BUG] Windows + Git Bash: tengu_cobalt_ridge experiment silently force-enables the PowerShell tool & "Shell: PowerShell", causing wrong-shell retries (token waste)
- [Bug] 2.1.158 (Windows): multiple tool/harness issues found in a stress-test
- Auto mode selected at plan approval persists after execution completes
- [BUG] Persistent server-side session billing phantom usage — cannot be killed from client
- opus-4-8 appears to re-fetch identical tool results after normal (non-error) returns — 2-3x vs opus-4-7
- [BUG] the cowork VM startup failure (HYPERVISOR_VIRT_DISABLED) is showing as a thread-blocking modal
- Enriched input signal for Claude Code - Eyeballs for non-verbal communication cues
- [MODEL] Opus 4.8 fabricated a present-tense user request and persisted on an invented task context
- [Bug] Claude responding in Japanese when prompted in Chinese
- [FEATURE] Allow users to opt out of post-compaction continuation injection via CLAUDE.md
- [BUG] agent stuck: git-rebase preference + auto-mode -> blocked force pushes -> requires human intervention
- [Feature] Add a user-facing toggle for spinner verbs / tool-use summaries in Claude Desktop
- [BUG] Headless mode (`-p --output-format stream-json`) breaks AskUserQuestion handling — model continues on assumptions instead of stopping
- Agent tool model parameter should accept full model IDs, not just tier names
- No way to set effort level per-subagent
- [Bug] v2.1.156+ regression: native Read/Bash return fabricated-but-plausible content, triggering agent confabulation cascades
- C:/Program Files/Git/schedule fails with "trouble connecting with your remote claude.ai account" — persistent, no public status incident
- [FEATURE]
- `claude --bg` code-writing sessions stall on permission prompts with no non-interactive way to reply (acceptEdits insufficient)
- Docs: clarify that `claude --bg` / background dispatch sessions do not carry `agent_id` in hook payloads
- Windows: background daemon keeps a directory handle after `claude rm`, blocking directory deletion
- [Bug] Auto-accept mode ignores directory containment check for home paths ending in period
- [BUG] `/insights` skill: AI-generated sections empty (returns {})
- [BUG] LSP client never sends textDocument/didClose, leaking language-server resources until OOM
- autoCompactEnabled does not trigger automatically despite being set to true
- /compact fails with 'thinking or redacted_thinking blocks cannot be modified' error when context is too long
- [Bug] Claude 4.8 model stuck in infinite loop executing spurious bash commands
- Wrong Charges of extra credits!!
- [Bug/Regression] Arrow key navigation skips multi-line prompt navigation
- [FEATURE] Per-banner "Don't show again" dismissal for MCP connection notifications
- [Bug] Session cost significantly higher than expected; parallel agent forking causing excessive API usage
- [Bug] File read errors not surfaced, causing hallucinations in tool use
- [Bug] Excessive token consumption on basic operations
- [Bug] Claude Code fabricated test results and pushed untested code to main branch (Opus 4.8 @ Effort = High)
- [UX] Add first-class false-positive reporting for safety guardrail blocks
- Ordered-list markdown in submitted messages gets auto-renumbered (e.g. '2.'/'5.' → '1.'/'2.'), clobbering intended numbering
- Permission/question dialogs do not render in Ctrl+O transcript mode (UI hangs indefinitely)
- [BUG] claude recommends adding non-existent directory to PATH
- [Bug] Serializer-complete rule violation in conversation state management
- [Bug] Auto-update failed - /doctor command produced no relevant output
- [BUG] Personal repositories not showing in Claude Code repo picker
- Files being truncated by edit in cowork with no warning and way below the 32kb limit
- [BUG] Opus 4.7 excess token usage in Windows - 7K tokens used 52%
- MCP: re-enumerate tools mid-session for remote-control sessions (mcp_reconnect works but is unreachable)
- [BUG] claude code harness sent "data has changed" system reminder with 10-days old date, claude code accepted it
- [Bug] Tool invocation fails due to extraneous characters before tool call
- [Bug] Agent ignores user instructions, commands, and skills
- `acceptEdits` mode does not extend to `additionalDirectories` (Edit/Write still prompt per-file)
- Bash sandbox (bubblewrap) corrupts `!` to `\!` in commands, making the sandbox unusable for agentic workflows
- Feature Request: Show usage stats in account dropdown menu
- Workflow run shows "Running 2032m" indefinitely after the agent has died; cannot be stopped via UI Stop button or TaskStop (Run ID / Task ID both unknown)
- Session title diverges between Desktop app and mobile/web — two independent title stores that never reconcile
- [BUG] Cowork MCP client silently drops unknown args and rejects missing-required locally, hiding the server's structured validation error
- [BUG] Cowork: large tool results spill to a file with a preview that truncates load-bearing fields (e.g. record IDs), encouraging fabrication
- [BUG] Cowork: uploaded context lands in the project mount, not mnt/uploads/
- [BUG] Slash command autocomplete should prioritize built-in commands over skills with the same prefix
- Benign `node -e` regex script false-flagged as "Fork bomb detected"
- Claude Code deleted client video files without confirmation — data loss on production work
- Managed/team-seat config bypasses MCP tool-search deferral — claude.ai connector schemas re-inject into messages every turn (CC 2.1.158)
- Claude Code context window not resetting — stuck at 1M instead of 200k on Pro
- Feedback: Claude should observe existing layout structure before adding fixed/static UI elements
- [Bug] Stray token "count" injected before tool calls causes parse failure in Claude Code
- Agent tool: `isolation: "worktree"` creates worktree in session repo, not target repo, breaking parallel cross-repo work
- [BUG] MCP: structuredContent overrides content field, losing actual tool response
- Intermittent tool-result corruption: stale/replayed output + sibling cancellation (v2.1.143)
- computer-use MCP: browsers reported "(not installed)" on macOS when Spotlight is disabled, despite LaunchServices having them registered
- Model autonomously deployed to production without user authorization
- [BUG] Cowork shows "yukonSilver not supported (status=unsupported)" on Windows 11 Pro with 11th Gen Intel i5-1155G7 — VM never initializes
- [BUG]
- Worktree isolation: Edit-tool path validation + drift-detection in task-notification payload
- advisor server-tool emitted but silently returns no result (stop_reason=end_turn) → model confabulates; regression from ~2026-05-28
- [Bug] Claude Opus 4.8 API returning increased hallucinations
- [Bug] Opus 4.8 hallucinating security incidents and fabricating evidence during long tasks
- [Feature Request] PostToolUse hook で output 文字列を改変 (REDACT) する機能の追加
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [BUG] Workflow harness retries indefinitely on HTTP 429 rate_limit — 97 agents, 2M tokens burned in 34 seconds ()
- Opus 4.8 Model stated multiple unverified claims as fact and fabricates specifics within a single session
- [BUG] Installer places binary in VS Code Snap directory, leading to silent uninstallation on Ubuntu
- [FEATURE] Plan mode — AskUserQuestion modal covers the agent's analysis and can't be minimized
- [BUG] Dispatch thread permanently stuck — server-side reset required
- Batched/parallel tool calls: one failure cancels all siblings; Edit old_string match failures (Windows CRLF)
- Bash command with 2>&1 causes ~12min spinner hang with token usage climbing during the stall
- [MODEL] Porting JS to TS Relies Too Much On Feedback From TSC
- Multiple profile directories (~/.claude-a, ~/.claude-b) require independent re-logins despite sharing the same account
- Bug: Remote Control via remoteControlAtStartup has no persistent indicator — startup notice scrolls off, nothing in status line / corner
- [FEATURE] PreToolUse hook: let "ask" decisions set the default-highlighted prompt option
- [BUG] Thousands of ghost "root-server" sessions regenerate on every launch, crashing the app — survive full uninstall/reinstall and all known data clearing
- [Feature Request] Improve readline keybinding compatibility for word navigation (alt-f/alt-b)
- [Bug] Model version inconsistency across devices with same account, requires reinstall
- Display `~/` instead of full `/Users/<name>/` in tool call path labels
- [Bug] Excessive token consumption from repeated echo tool calls
- [Bug] Terminal output lag and duplication in Claude Code TUI rendering
- [MODEL] Claude ignored instructions/configuration (also: Claude made incorrect assumptions)
- Parallel tool calls: one benign tool error cancels all sibling calls → agent misreads as broken shell and flails
- [FEATURE] ature request: voice (text-to-speech) output in Cowork and Claude Code
- [FEATURE] Dynamic Workflow Feature List
- Claude Code VS Code extension forces 1M context on Pro plan - Windows
- MCP gateway WAF blocks legitimate tool-call bodies (SQL/shell/config text)
- [Feature Request] Selective Context Compaction for Partial Message Pruning
- Unprocessable image in context burns credits with no recovery path
- Image attachments silently fail to upload when message has no text
- [Feature Request] Sidebar: nest sub-folders within the same Git repo as separate project blocks
- [BUG] Opus deleted entire repository without realizing it (corrupt tool call?)
- [Bug] Agent stops responding after calling advisor tool
Resolved issues (422)
- [BUG] Auto-compact not triggering before 1M context limit error in Cowork mode
- Tool calls leak as raw text ('call' + args) instead of executing — frequent on Opus 4.8
- Opus (CLI) took down a node's DNS by deploying a k8s LoadBalancer on :53, after dismissing on-screen evidence the port was the host resolver
- Hook stdout shown + suppressOutput:true ignored in 2.1.158 (regressed from 2.1.157)
- [BUG]
- Loading animation messages (Booping, Grooving, Zigzagging) are distracting/unprofessional
- [BUG] Invalid Oath - Claude Code
- [FEATURE] Support Cmd+V image paste from clipboard
- [FEATURE] ! shell mode: filesystem tab completion
- [Bug] Anthropic API Error: 529 Overloaded
- [FEATURE] confirm before deleting session
- [BUG] Windows: Grep tool leaks rg.exe + conhost.exe processes (~2000 zombies / 14 GB RAM in long sessions)
- [BUG] Restore from archive deleted sessions instead of restoring them
- [Feature Request] Show timestamps for each message
- [Bug] Anthropic API Error: Server rate limiting on concurrent requests
- [Bug] Anthropic API Error: Server rate limiting during request execution
- [Bug] Anthropic API Error: Server rate limiting despite normal usage
- [FEATURE] Add context window usage indicator and warning before auto-compaction
- [BUG] Cowork downloads Linux ELF binary instead of macOS binary on macOS Sonoma 14.8.7 — exit code 132 (SIGILL) on every session
- [BUG] Claude Desktop reinstall destroys Code chat history (transcripts + Recents) while regular Chat history, project files, and memory all survive
- Model fails to read/reference CLAUDE.md contents despite being loaded in context
- [remote-control] Input from mobile app/browser not reaching host session — output works fine
- Paid yearly subscription silently downgraded to Free with no user action
- [BUG] Default git workflow in the system prompt was pushed without context or consent
- [BUG] claude rm/stop/reap SIGKILLs background session tree without SIGTERM grace, orphaning git index.lock and similar
- [BUG] Tailscale via Homebrew CLI + Mac App Store GUI, both Macs on macOS, Cowork blocked by VPN detector despite Tailscale being a mesh VPN with no traffic interception
- [BUG] Claude Code on the web: autofix sessions stopped receiving bot PR review comments (Copilot/Codex)
- [Bug] Image Processing Failure: Unable to Parse Crop Phenology Color Pattern
- [BUG] Korean (Hangul) IME: last syllable dropped on space/punctuation in input
- [BUG] CCR routine runs stuck at "setting up a cloud container" — stop button doesn't clear status
- [Bug] Transcript corruption from unsafe tool_result/thinking block separation during compaction
- [BUG] Bug: 400 error "thinking blocks must be passed back" with third-party compatible endpoints after upgrade to 2.1.152
- [FEATURE] GUI project/session switcher in Claude Code Desktop
- [Bug] Claude 4.6 Model Instability: Frequent Instruction Drift
- [Bug] Auto-update failed with diagnostic error
- [BUG] [Bug] Cowork Desktop only shows Sonnet 4.6 (1M) — no standard 200K option, triggers usage credits error on Max plan
- enhancement: first-class subagent policy primitive — explicit in-prompt rules still get rationalized past
- docs: clarify that project CLAUDE.md / AGENTS.md are not inherited by subagents
- Claude Code sessions inconsistently sync to Claude iOS app — some sessions visible, others not
- [BUG] Cowork can't mount paths inside Dropbox on macOS.
- [BUG] Cowork sandbox FUSE mount serves stale inode for entire session — silently corrupts bash git commit; now also corrupts .git/config on every Windows-side gh pr merge
- [Bug] /exit with worktree generates invalid resume command
- Org-configured plugin marketplaces not synced to local CLI on login
- [BUG] Claude Code VS Code extension truncates HTML files on save (LF→CRLF rewrite chops trailing scripts)
- Claude Desktop doesn't pass --dangerously-skip-permissions to child claude processes even when bypass is opted-in
- [BUG] "Allow bypass permissions mode" setting enabled but sessions always start in Accept Edits mode
- [BUG] `isolation: "worktree"` on Agent tool creates worktree but doesn't bind subagent CWD or rewrite parent-path references
- Per-project plugin enable/disable via .claude/settings.json
- Auto-mode classifier blocks read-only prod-DB reads on user-owned infra with no in-session escape hatch
- /ultrareview crashes during Find phase before producing findings
- [FEATURE] Expose image path/data to MCP tools when backend model doesn't support images
- Auto-rename active session to match current git branch (incl. parallel-agent / worktree workflows)
- [Feature Request] Add lighter styling options for prompt area in fullscreen mode
- [BUG] Cowork: 403 Request not allowed on every request — Pro plan
- [BUG] Write tool gate denies path that permissions.allow glob clearly matches
- [Bug] Remote Control injects claude.ai account MCPs into CLI session without opt-out
- [Bug] Agent fails to distinguish taxed vs untaxed prices despite clarification attempts
- Multi-root workspace UX: show active CWD and make root switching obvious
- skill-creator: Windows incompatibility in run_eval.py — select.select on subprocess.PIPE fails (WinError 10038)
- [BUG] Files open in same panel/sidebar as Claude Code chat instead of main editor
- [BUG] Auto-updater rewrites ~/.local/bin/claude as self-referential relative symlink → ELOOP (recurrence of #45260, on 2.1.152)
- Feature: sticky input prompt while scrolling output
- [FEATURE] Add option to disable/lock the delete button for chat history in VS Code
- You've hit your session limit · resets 5:30pm (Asia/Jerusalem)
- [FEATURE] CLI flag / settings.json key to pre-select session-resume mode (auto-restart use case)
- Claude Code 2.1.149 crashes on launch — Bun v1.3.14 SIGILL/segfault on Intel macOS (x64)
- [FEATURE] Recursive repo discovery for claude agents
- Update stuck: empty version file prevents re-download
- [BUG] /account shows 'Not authenticated' in older windows after another window logs in/out, while Usage bars still populate
- ultrareview: two consecutive crashes on same PR ("Review crashed before producing findings")
- effortLevel "max" in settings.json not respected on session start — falls back to xhigh
- [FEATURE] CLIでCLAUDE Codeが次の回答（本来は私の回答）を予測して表示する文言をそのまま使いたい
- [BUG] tui: "default" setting is ignored when attaching to a session from agent view (claude agents)
- Desktop app v1.9255.2 (CCD 2.1.149): Bypass Permissions mode fails — session not launched with --dangerously-skip-permissions
- [FEATURE] Accept & Reject buttons
- [BUG] "you've been signed out" even if I have not error message
- [BUG] After /compact, statusLine receives one tick with used_percentage=0 and input_tokens=0
- billing is unfair and poorly documented
- Claude Desktop's Info.plist missing TCC usage strings, blocks all EventKit-based MCP servers
- [Bug] Scrolling disrupted by output while reading chat history in iTerm2
- [Feature Request] Periodic silent execution primitive for autonomous agents
- [Bug] Unexpected token consumption without user interaction or API calls
- [FEATURE] Add support for devcontainers to Claude Desktop (for code tab within Desktop)
- [BUG] claude code can't access external projects if running via sandboxed environment
- [Bug] Claude Code exits silently without error logging on Windows PowerShell
- [BUG] "Instalação no Windows PowerShell retorna vazio"
- [BUG] Per-prompt cap on claude-opus-4-7 silently dropped ~2x around Apr 27, 2026
- [BUG] double billed for claude max plan 2 months in a row and then account degraded to FREE PLAN
- [FEATURE] Group diff files by directory with collapse-by-default option (Desktop)
- Sandbox blocks macOS SystemConfiguration API, crashing Rust-based CLI tools (Codex CLI)
- [BUG] Claude Desktop Cowork: "Check your Claude in Chrome account" error persists despite correct account, working native messaging, and active socket connection
- [BUG] AskUserQuestion returns 'undefined is not an object (evaluating H.map)' on subsequent calls within same turn (2.1.122)
- remote-control: 403 token expiry kills all active sessions instead of refreshing
- [BUG] Cowork: workspace directory names should not contain spaces (causes build/tool failures)
- [BUG] Plugin install does not seed userConfig.*.default into settings.json — surfaces as skill !-context bash failing "Contains expansion"
- [Bug] System rules not applied during code generation or tool execution
- [Feature Request] Improve permission prompts: reduce noise for compound commands and repo subdirectories, delegate script review to subagent
- CC tasks disappear from TUI after /compact (session ID not preserved)
- /ultrareview consumes free run on Find-phase timeout failure for oversized diffs
- [FEATURE] customizable terminal input box cursor shape
- Feature request: fast-mode dispatch for Agent() subagents (model=opus-fast or fastMode flag)
- [BUG] Claude input overlay remains visible when editor group is maximized in VS Code
- Project-root `config/` directory auto-deleted when Codex plugin installed (data loss)
- Edit tool results in conversation history include full originalFile, causing context pollution and model confusion
- [BUG] Schedule entry missing from desktop sidebar (present on web)
- [BUG] Write permission allowlist ignores relative file_path that resolves to absolute-path rule target (still broken in 2.1.122 — see #29098 / #37220 / #38391)
- claude --resume <session-id> always returns 'No conversation found' even though session files exist
- CLAUDE_PLUGIN_ROOT env var not always set when invoking plugin hooks
- Bring back custom folders in the desktop sidebar
- [Bug] SessionStart startup hook error: ToolUseContext is required for prompt hooks
- Add keyboard shortcut to copy inline response to clipboard cleanly
- Feature request: set distinct CLAUDE_CODE_ENTRYPOINT (or similar env var) for Agent-spawned sub-agents
- full agent-discipline harness built on Claude Code primitives
- [Bug] Rate limit error on /ultrareview consumes free attempt quota
- Terminal becomes unresponsive when running grep on external SSD
- Telegram channel MCP server disconnects after 5s, drops messages
- Telegram plugin: inbound notifications not reaching session
- stdio MCP server killed every ~5 min when JSON-RPC ping response misses timeout window
- [BUG] autoMemoryDirectory in settings.local.json is ignored - system prompt still hardcodes global memory path
- [BUG] API ConnectionRefused and Image Recognition broken after upgrading to 2.1.122 with MiniMax
- [BUG] Claude Code OAuth breaks when workspace setting "Allow creating new API keys in default workspace" is disabled
- [DOCS] Incorrect npm package in code block
- Routines: show scheduled task list in sidebar without navigating away
- [BUG] Claude Code stops before my limit rates
- [Bug] Model performance regression: Claude Code output quality degraded
- [BUG] JetBrains plugin treats claude.exe path as directory for IDE lock file on Windows
- [Bug] Claude Code fails to resolve CSS image alignment issues
- [BUG] desktop app freezes on every session after the last update this morning. Totally useless
- [BUG] Marketplace plugin sync permanently blocked by PNG compression ratio false positive — stale server-side cache prevents recovery
- claude code not connecting to chrome browser
- [FEATURE] Voice dictation: option to reduce prosody-driven punctuation insertion
- [Bug] Prompt cache unexpectedly collapses mid-session on Opus 4.x
- [BUG] slow
- PermissionRequest hooks silently deny in /remote-control when registered with no permissionDecision output
- /ultrareview cloud session fails with `CLAUDE_CODE_OAUTH_TOKEN not set`, then silently times out at 30 min — consumes free runs
- [Bug] Anthropic API Error: Failed UltraReview Request
- [Bug Report] you ai is completely retarrded give me my refund asap
- Commercial spam pattern by @junaidtitan (Junaid Q / Ruya-AI) across 50+ claude-code issues
- [FEATURE] Allow Claude to programmatically rename conversations via a tool or CLI command
- Cowork sessions don't inject user_config env vars into MCP server processes
- [BUG] claude has been a complete idiot that last four days!!
- Process crashes with exit code 3221225477 (0xC0000005 STATUS_ACCESS_VIOLATION) on Windows ARM
- Feature request: ability to migrate/transfer a Claude Code session to claude.ai browser chat
- Claude Desktop freezes completely when Computer Use (Teach Mode / Claude in Chrome) is initiated
- [Bug] Model instruction following degradation in Claude 4.6 & 4.7- ignores explicit code replacement directives
- remoteControlAtStartup: true in settings.json does not auto-enable remote control on new sessions
- [BUG] Cowork Save Skill fails with INVALID_PATH when Mac username contains a space
- Skills pane shows empty state during cloud-sync, indistinguishable from "no skills configured"
- VS Code chat-panel webview ignores leading-slash workspace-rooted markdown links
- Chinese (Simplified) UI Localization for Claude Desktop App
- Feature: Version-pinned model parameter in Agent tool
- [BUG] Plugin hooks.json entries with shell-spawning command silently skipped on Windows; node-spawning entries trigger normally
- [BUG] Claude Desktop App crashing each time I select 'Code'.
- [MODEL] opus 4.7
- [BUG] Cowork session crashes immediately on macOS 26.3.1 — VM kernel panic "Attempted to kill the idle task!"
- [FEATURE] In Desktop SSH sessions, allow WebFetch to run on the local host instead of the remote SSH target
- Claude Code enters fix loop on mobile image/object-fit issue — 5+ attempts without resolution
- [Bug] Excessive API calls causing unexpected billing on failed requests
- [BUG] Code block contrast too low in Claude Code web UI (3+ days)
- Allow suppressing or customizing the PR number in the built-in mode indicator
- [FEATURE] Allow opening the right-side panel (terminal/navigation) before starting a conversation
- [FEATURE] [Windows] Cannot paste clipboard images (Ctrl+V) in Claude Code desktop app
- Agent loops on Read tool when saving memory — re-reads same file repeatedly
- [BUG] Monitor Tool Notification Flooding in Poll Loops
- [Bug] Claude Code fails to read project documentation and provides misleading information about system access
- [FEATURE] Allow worktree branch renaming mid-session for meaningful git history
- Claude Code desktop: "Auto-fix CI & address comments" doesn't dispatch to local session
- [FEATURE] Token limits make sustained project work impractical on Pro plan
- API error and than get rate limit error without any work output
- feat: honor MCP CallToolResult _meta.cache_hint to opt out of prompt caching per result
- GitLab plugin: support configurable base URL for self-hosted instances
- ultrareview (no-arg): misleads users into thinking it reviews the full codebase
- Sync Claude Code (VS Code) conversations with Claude Web & Mobile App
- [MODEL] opus4.6
- [BUG] Orphaned --resume process consumes tokens silently after WSL reconnect + extension update
- [BUG] IDE extension bypasses .claudeignore and settings.json deny rules (Auto-attach context leak)
- [FEATURE] Persist session history with IDs and dates for easy resuming
- [Bug] Claude Code launched duplicate bot instance without checking for existing deployment
- [Bug] /ultrareview misleading behavior: reviews only branch diffs, not full codebase
- [BUG] [ultrareview] Remote review returns empty findings ([]) on large-scope repositories
- [BUG] Shift+Tab accept-all mode still prompts for Read/Edit on every file.
- Ultraplan: stream timeouts, cache_control on empty blocks, teleport failure
- Feature request: deep-link to specific Cowork scheduled task + propagate task identity to MCP tool calls
- [Bug] /mcp command produces no dialog or output
- Skill tool feedback loop: getPromptForCommand body re-injected as isMeta after every model-invoked Skill call
- CLAUDE.md instruction adherence significantly degraded since ~April 4
- Stats dashboard streak broken despite valid session data on disk
- [BUG] Cowork shows "Virtualization is not available" on Windows 11 Pro 24H2 — all Hyper-V services running, vm_bundles never created
- Agent teams: tmux socket unlinked while server alive — ghost servers and leaked tmux-spawn scopes
- [BUG] Linear MCP tools appear as "deferred" and fail to work reliably across sessions despite server showing Connected
- Spurious ~/.local/bin PATH warning on Windows
- [BUG] Usage drains in seconds of Opus usage
- You've hit your limit · resets 1:30am (Asia/Dhaka)
- You've hit your org's monthly usage limit
- [BUG] Mac Desktop: disappearing sessions bug -> React
- [BUG] Multi-line paste fails in Ghostty 1.3.x with TERM=xterm-ghostty (works after switching to TERM=xterm-256color)
- [BUG] Cowork "Virtualization is not available" on Windows 11 Pro 25H2 — all virtualization features enabled, VM never initializes
- [MODEL] CLI
- I appreciate the feedback, but I should clarify: I'm Claude, an AI assistant made by Anthropic. I'm not actually "Claude Code" or a GitHub issue tracker.It sounds like you're sharing thoughtful product feedback about Claude Code's interface and features
- Subagent worktree: non-idempotent mkdir + process handle leak (Windows 11)
- [BUG] /ultrareview failed TWICE
- Support Chinese (Traditional/Simplified) UI language
- [BUG] v2.1.118 Reconnect action does not reach mcpServers entries added by direct ~/.claude.json edit
- [FEATURE] Allow additionalContext to specify a /context category
- [BUG] Claude Code never invokes tools, only returns text responses (Windows 11, v2.1.123, acceptEdits mode)
- Usage allowance shared across two separate Anthropic accounts in parallel VS Code Claude Code sessions
- Plan pickup is BROKEN. You told users a new session resumes. It DOESN'T.
- [Claude in Chrome] window.print() dialog freezes renderer; need silent "save page as PDF" tool
- [BUG] MCP tools/call returns "Tool call failed: 500" against server that responds with HTTP 200
- C:/Program Files/Git/stats panel undercounts Active days vs local session JSONLs (24 on disk → 8 reported)
- Image drag-and-drop broken in tmux (regression)
- [MODEL] ESTAFA PLAN PRO 20 usd. Alcanzando el limite diario con un excel de 40 registros
- [FEATURE] AskUserBash tool — let agents prompt for secrets without exposing them to the model
- [Bug] Incorrect token usage calculation in new session
- Claude Code fails game developers on specialized platforms — BGA case study
- [UX] "Fork from here" button placement — should be under Claude responses, not user prompts
- You've hit your limit · resets 7:10am (Australia/Sydney) - while I just barely started work after 1 day break
- Scheduled wake-ups fire into wrong session when two sessions share a project dir
- [BUG] Claude Code fails to follow established process rules despite repeated correction
- Double-Esc and `/mcp` hang when an MCP server connects to a filtered port
- [BUG] Ultrareview consumed free quota despite returning empty findings
- [BUG] # shortcut does not write to CLAUDE.md
- [Bug] Connector marketplace OAuth fails for multi-region SaaS partners with region-specific authorization servers
- [Bug] RemoteTrigger.run MCP action returns HTTP 400 — trigger_id sent in body instead of URL path
- [BUG]
- Sessions disappear from recent conversations on desktop app restart (Windows)
- [Bug] Conversation scroll-up shows duplicate content and prompt field gaps
- Terminal resize causes duplicated/corrupted output in tmux (regression in 2.1.116)
- [Bug] Duplicate conversation output lines displayed in terminal UI
- [Bug] Brainstorm mode displays Greek letters instead of numeric options
- [FEATURE] Mobile app: Edit/Write tool prompts don’t let user preview the diff before approving
- [BUG] AskUserQuestion's "preview feature" doesn't allow "Type something" as an option
- Chrome extension bridge rejects valid OAuth token: Invalid token or user mismatch
- [Bug] Claude Code ignores provided documentation and generates incorrect information
- [Bug] Edit tool fails on Windows with CRLF line endings: Read output mismatches file content
- [BUG] Self-authored escape valves in subagent prompts override checklist constraints
- [BUG] IME composition Enter key triggers submission on claude.ai/code (Chrome)
- Interface switching should not destroy conversation context
- [Desktop][macOS 26] Code tab crashes — rootfs.img redownload loop saturates disk writes
- Native CLI 2.1.152 consumes 8 GB+ RAM on regex with bounded repetition over minified single-line input
- [Feature] Expose task/tool context in Notification and Stop hooks
- [Bug] 2.1.158 (Windows): multiple tool/harness issues found in a stress-test
- [BUG] Image paste (Alt+V) always fails under WSL when login shell is fish — clipboard command run via $SHELL, which rejects "$(...)" in command position
- No way to set effort level per-subagent
- Channel splash shows 'plugin not installed' for all plugin-type channels despite channels working correctly (v2.1.145)
- [BUG] Agent Teams (in-process): unbounded team-lead inbox stalls teammate→lead message surfacing at ~1MB / 1000+ msgs
- [BUG] Severe over-triggering of safety guardrails — Claude hallucinates policy violations in completely harmless prompts
- [Bug] Tool invocation fails due to extraneous characters before tool call
- [BUG] Regression in 2.1.158: claude spirals into redundant/invented tool calls during file reads (2.1.157 clean)
- advisor server-tool emitted but silently returns no result (stop_reason=end_turn) → model confabulates; regression from ~2026-05-28
- Automatic light/dark theme selection?
- Claude Desktop writes malformed registry Run entry (nested escaped quotes) - crashes Windows Task Manager and other Run-key parsers
- [BUG] Claude Desktop — Personal plugins panel renders list but is entirely non-interactive (macOS, v1.9255.2)
- [BUG] DesktopvsApp
- [BUG] FUCK CLAUDE
- [BUG] in security_reminder_hook.py , CHANGELOG.md , gitutil.py , llm.py , feed.xml
- Claude Desktop (Windows MSIX) auto-update silently fails with 0x80073D02 while app is running
- [Bug] Auto-classifier incorrectly blocking legitimate PR reviews as false positives
- [Bug] Anthropic API Error: Usage credits required for 1M context window
- [BUG] Claude giving free extra design token usage while restricting paying customers.
- [BUG] Cowork macOS sidebar overlay persists through reinstall - blocks conversation content
- [Bug] /agents command freezes on desktop CLI, unable to create new agents
- [BUG] IME pre-edit buffer cleared by UI redraws during candidate selection — McBopomofo / 小麥注音, macOS, iTerm2
- [BUG] .claude.json and backup files disappearing without user action — twice in 48 hours
- Teammates don't inherit bypassPermissions from parent session
- Add setting to disable ide_opened_file auto-context from VSCode extension
- VSCode Extension: Add setting to disable automatic IDE selection/context injection into conversations
- /mcp dialog hangs for several seconds when an HTTP MCP server has stale OAuth tokens but reports "Connected"
- [Bug] Claude Code commits large bulk changes across multiple similar files without user confirmation, triggering abuse detection systems
- [BUG] Conversation mode crashes when connected to CarPlay
- [FEATURE] export: add --format flag (md/json) and --last N selective export
- [BUG]
- [BUG] Auto-pilot after compaction
- [BUG] Rapid quota consumption in Claude Code with Opus 4.7 1M context — minutes-scale 90%+ burn from few prompts
- [BUG] Cowork VM root partition 86% full on fresh image — only 940 MB free for user work
- PreToolUse hook input missing agent_id for tmux/pane-based teammates
- Push notifications not delivered on Android (Samsung S23) despite Remote Control showing Connected
- Telegram channel plugin MCP server not spawned with --channels in v2.1.101
- [BUG] Artifact API proxy strips mcp_servers parameter
- [BUG] SendMessage no longer available
- [BUG] Telegram plugin's MCP server hijacked by other Claude Code sessions, ignoring `--channels`
- Duplicate console output across multiple independent projects (macOS)
- [Bug] Duplicated output during inference
- [FEATURE] make tools like ripgrep and fdfind a first class citizen
- [Bug] Claude Code not retaining task list items and skipping tasks during execution
- [BUG] Claude API with Claude Code
- MCP tool calls display generic 'Called <server>' instead of tool name and params since ~2.1.105
- claude mcp add --scope user writes to a location claude mcp list does not read
- ultrareview: free run consumed when server-side rate-limited (run crashed before producing findings)
- [BUG] Pro plan session limit exhausted on single non-coding audit prompt
- [BUG] UltraReview timeout/crash consumed ~$37 in credits with no output returned
- [BUG] /plugin "Will install" iterates hooks path as char/index list
- [BUG] .skill writes from channels still does not work with DSP
- Feature request: let hooks invoke slash commands (e.g. /compact) via hookSpecificOutput
- Claude in Chrome: CLI never connects to native-host bridge socket (macOS, v2.1.121)
- [BUG] Claude Code agent ignores CLAUDE.md / project directives, leading to real-money loss
- [BUG] In-process teammates lack extended thinking — capability gap with tmux teammates and team lead
- Sandbox bind-mounts /dev/null device nodes in working directory, breaking Python build tools
- Custom MCP HTTP connector handshake stalls after notifications/initialized and GET /mcp — "Couldn't reach the MCP server"
- [FEATURE] Computer Use support in Claude Code CLI on Windows
- Plugin slash-command autocomplete drops namespace prefix on accept
- [Feature Request] Add option to suppress follow-up prompts
- [Bug] Conversation state corruption with ID 88f9f0c4-b087-45a2-8701-007d1c88f1de
- [Opus 4.7] "Claude failed to flag the well-known WooCommerce shop page hijack despite multiple iterations of the same problem; cost ~24 hours of debugging"
- [FEATURE] Extend deferred tool loading to subagents — refile of #31623
- [BUG] CronCreate: lacks predicate-gated execution and context isolation, leading to large unintended costs on repeated trivial checks
- Edit tool silently normalizes `\uXXXX` escape sequences in old_string, breaking exact matches
- [Feature Request] Add support for autonomous web scraping agent for outbound lead generation
- Agent tool's `model` override silently ignored — fork always runs as parent's model
- [BUG] Cowork Linux sandbox fails with HYPERVISOR_SERVICE_ERROR (0x80370102) on Citrix XenServer-based VDI environment
- Allow --resume to accept session name as an alternative to UUID
- [FEATURE] Configure language for auto-generated session/chat titles
- [BUG] crash-on-"hi" behavior
- Sessions bar: renaming a session via pencil icon reverts immediately on blur/Enter
- [Bug] Unexpected usage limit reset cycle change from Monday to Friday
- Edit tool: InputValidationError about required parameters missing, intermittent, retries succeed
- Multiline paste broken in v2.1.105 on Windows (mintty/Git Bash)
- Read tool: error message should mention offset/limit parameters
- Plan mode conflates 'format my document nicely' with 'execute the plan'
- [BUG] settings.json stale-write clobber silently disables installed plugins (continuation of #27941)
- [FEATURE] PreAgentSpawn hook: executable cost policy for sub-agent dispatch
- [BUG] VS Code Extension - Current prompt & file visibility do not persist session switches
- [FEATURE] Usage display should surface cache read and write token consumption
- [FEATURE] Support for Profile and/or Project filtering on /insights command
- [BUG] ScheduleWakeup paired with Agent (subagent) dispatch is redundant — Agent already emits completion notification
- Dispatch messages show as read but never execute after rate limit (macOS, v1.1.9493)
- [OPUS4.7] Read tool returns different content for ~/.claude/settings.json across calls within the same session — effective-view rendering or bug?
- [Feature Request] Restore recent routine chats sidebar visibility for quick access and approval status
- Proactive `/schedule` and `/loop` pitches feel intrusive — please add a global opt-out
- Agentic sessions: fabricated confidence and approach-thrashing instead of finding prior art
- CUA save_to_disk screenshots not accessible from filesystem
- [Bug] /ultrareview crashes without producing findings, incorrectly consumes free quota
- [Bug] autoMemoryDirectory option not functioning after upgrade to v2.1.121
- [BUG] Scheduled triggers stuck at "Coalescing..." / "Processing..." — sessions never execute
- Desktop app silently creates settings.json with effortLevel: high on first launch, causing unexpected API cost spike
- Auto-update deletes session .jsonl files, wiping token usage history and conversation data
- [BUG] Claude constructs &&-chained command with ||-tolerant loop + trailing rm -rf, swallowing failures and wiping working tree
- /loop skill has no cancel mechanism — loops run until session ends
- [BUG] macOS App - Claude Code Failing to Retriver of Session & Failing to Initiate new ones
- [Bug] Anthropic API Error: Cyber Verification Program blocks legitimate periodic monitoring tasks
- Is there a way to configure the `persistHookOutput` 10,000-character threshold?
- [BUG] Scheduled tasks fire on cron but are not visible in Desktop UI (1.5354.0, Windows)
- [Bug] Loop command repeats identical responses instead of progressing to next steps
- [BUG] Claude Code hangs with local Ollama on a trivial prompt, while direct /v1/messages works
- Bug: Claude Code Installer Improperly Modifies ~/.zshrc Without Newline Separation
- claude setup-token outputs OAuth token wrapped across two lines, breaks copy-paste
- Feature request: disable suspiciousPathGuard for bypassPermissions users
- [Bug] Unexpected high usage consumption on Claude Max plan (using claude code)
- [BUG] Uppercase German umlauts (Ä, Ö, Ü) silently dropped in chat input on Linux/WezTerm
- [BUG] Thinking stalls exceeding conversation cache TTL
- Claude Code modified 1,008 user files without permission during 'test' of installed automation
- [FEATURE] Allow project-local task list directory
- Worktree dialog: selecting branch before checking "create worktree" reverts branch to default
- [BUG] spinnerVerbs, SpinnerTipsOverride, and spinnerTipsEnabled settings are ignored
- [BUG]
- [BUG] Unexpected value(s) `effort-2025-11-24` for the `anthropic-beta` header. Please consult our documentation at docs.anthropic.com or try again without the header.
- [Bug] Subagent inherits parent's context window size instead of model default
- [BUG] OAuth error: Request failed with status code 400
- [BUG] Cowork: Session init fails to mount scheduled task config folder outside project mount after macOS Desktop upgrade
- [BUG] Agent-Teams sub-agents with model: opus cannot call SendMessage / TaskCreate / TaskUpdate etc. despite frontmatter declaration (tmux backend)
- [Feature Request] Live theme detection for macOS auto-switching appearance
- Claude no longer generates PR title/body summary on "Manually create PR"
- [Bug] /ultrareview function crashes on execution
- [BUG] Claude reads rules but don't follow it.
- [Bug] Claude Code crashes on initialize payload receipt via SSH remote in v2.1.121
- Support categories/groups in marketplace.json for plugin organization
- [AskUserQuestion] Two minor UX issues: Enter key doesn't submit "Other" field, response renders inside a code block
- [Bug] Anthropic API Error: Server rate limiting prevents work across multiple Pro accounts
- [BUG] False "Hook Error" labels cause Claude to prematurely end turns (exit 0, no stderr, valid JSON)
- Auto-update removes user symlinks from ~/.claude/skills/
- [BUG] Claude in Chrome (Cowork) operated unintended device's Chrome in multi-device environment via Parsec
- Agent denies authorship of its own in-session destructive commits, using project's no-attribution rule as cover
- remoteMarketplaceClient rejects all third-party plugin sources: 'External plugin sources are not yet supported'
- Full skill list re-injected on every tool call — token bloat, performance, cognitive load
- [Bug] Missing "Pay Now" button when purchasing API credits
- Monitor tool and run_in_background Bash exit 1 spuriously when polling `gh` in a loop
- [BUG] The JSON schema for .claude/settings.json is outdated
- [BUG] Claude Code crashes on Windows Enterprise - EPERM: operation not permitted, uv_spawn 'reg'
- After compaction and subagent return, Main agent set's itself from Sonnet 4.6 to Opus 4.7 without user interaction
- macOS: zsh -c -l Bash-tool wrappers leak post-completion in run_init_scripts → execif, idle Ss with no CPU
- Vercel plugin: false positive "Use named exports" error on page.tsx / layout.tsx files
- claude --resume fails with 400 "text content blocks must be non-empty" when transcript contains image-with-empty-caption
- [MODEL] Opus 4.7: elevated unforced errors on 2.1.121 — three back-to-back hardware-debug sessions, "false verification" pattern correlated with CLI version bump
- [Bug] Model exhibits inconsistent reasoning, instruction non-compliance, and context degradation in extended sessions
- Plugin PreToolUse hooks silently override `bypassPermissions` / `acceptEdits` with no indication to the user
- [BUG] Report: Claude Code repeatedly violates loaded workflow skill, escalates hallucination across correction rounds
- Vim mode: support system clipboard registers ("*, "+) or clipboard=unnamed for cross-prompt yank/paste
- setup-token: parallel `claude -p` spawns sharing CLAUDE_CODE_OAUTH_TOKEN see ~5-10s server-side serialization tax (different tokens unaffected)
- Plugin skill `context: fork` + `agent:` frontmatter not honored in 2.1.112 — runs inline instead of dispatching to subagent
- Context window detection fails for third-party Anthropic-compatible providers
- [BUG] [Max plan] Weekly usage limit reset day shifting forward each week (Wed → Thu → Fri)
- Documentation ambiguity: CLAUDE.md concatenation order and priority
- [BUG] Claude Code claiming false Usage Policy violation when scanning emails
- [BUG] Weekly reset anchor shifted Thursday → Saturday in ~12hr window, Max 20x, no user action (Apr 28-29 2026)
- [FEATURE] Higher Subscription Tier (MAX x30) or Bulk Upgrade Option for users from Ukraine
- [BUG] `claude --upgrade` downgrades Claude
- Expose organization API credit balance to statusLine scripts (and via Admin API)
- [FEATURE]
- Bash permission walker reports "Unhandled node type: ;" and forces prompt on subshell-form commands
- [Bug] Unexpected high token consumption with non-Claude subagents and orchestrator models
- [Feature Request] Add localization support for non-English languages (Japanese, etc.)
- [BUG] claude behavior problem / skill-creator issue
- pr-review-toolkit: review-pr skill passes filenames not full diff to agents, causing hallucinated changes
- macOS: PhotoKit access blocked — Claude Code missing photos-library entitlement and NSPhotoLibraryUsageDescription
- [FEATURE] Embedded browser tools should resolve .local mDNS hostnames and inherit host OS network stack
- [BUG] Shift+tab has no action anymore
- [BUG] Weekly usage limit reset day drifted Thu → Sat on Max (5x), with timer rebase event
- Plugin autoUpdate doesn't update installed_plugins.json, leaving bundled hooks pinned to stale installPath
- Plugin loader uses wrong installPath — takes [0] from installed_plugins array instead of filtering by current project
- [FEATURE]
- [BUG] Claude Code crashes or fails with EACCES when connected via SSH to fresh Ubuntu server
- Feature request: auto-enter plan mode for slash commands / sub-agents
- [BUG] Claude Code desktop - Usage limit reached - while session reset
- Chat: add beginningOfBuffer / endOfBuffer actions for multiline prompts
- [BUG]
- Surface MCP server notifications/message as conversation interjections
- Claude not following instructions and jumping rails.
- [BUG] Cowork scheduled-task system silently dropping fires (recurring crons skipped on weekends, now bleeding into weeknights)
- [BUG] Setting only disableAutoUpdates locks Cowork and third-party inference panes as org-managed (macOS com.anthropic.claudefordesktop)
- [FEATURE] ature request: voice (text-to-speech) output in Cowork and Claude Code
- [Bug] Terminal output lag and duplication in Claude Code TUI rendering
- [BUG] Cowork hangs on "Starting up…" indefinitely + every message crashes — Windows 10 Pro, Desktop 1.8555.2.0