Digest / June 16, 2026
Claude Code Issues Digest — June 16, 2026
228 new issues filed ·
322 resolved in anthropics/claude-code (UTC day).
New issues (228)
- Intermittent malformed tool_use: model emits bare <invoke>/stray tokens instead of valid tool-call syntax
- [Bug] Claude 3.5 Opus: Excessive recency bias in task planning
- [Bug] Sessions disappearing from resume list and Opus 4.8 rate limiting errors
- Clickable file links with spaces in path (%20-encoded) don't open in VSCode extension
- Malformed tool calls: stray 'court' token + dropped antml namespace prefix on tags
- [BUG] MCP 403 insufficient_scope step-up authorization is dropped and misreported as "token expired"
- 2.1.178: native team-management tools TeamCreate / TeamDelete no longer surfaced (regression from 2.1.177)
- [BUG] Fable 5 fabricated a non-existent prompt-injection and self-sabotaged a local task (~47 min stall)
- [Bug] Performance degradation - slow response times
- [BUG] Claude Code commits without being asked, repeatedly ignoring CLAUDE.md rule
- [Feature Request] Filter worktree sessions to show only user's sessions
- [Bug] Insufficient Input Validation for Model Responses
- [BUG] Sonnet subagents fail from 1M context sessions with “Extra usage is required for 1M context” on Max 20x
- Fast mode billing bug
- feature-request-claude-code-link-resolution
- Desktop app: newly-created project (git repo nested inside an already-open parent folder) never appears as its own sidebar section
- PTY/file-descriptor leak in Claude desktop app exhausts all system pseudo-terminals (posix_openpt → ENXIO)
- [BUG] [VS Code Panel] Scroll does not work in conversation area — input box retains permanent focus
- [BUG] Options dialog covers response text in Cursor terminal (widescreen, wide option lists)
- Local .mcpb install silently fails (no dialog, no error) when any bundled file exceeds ~108KB
- [Bug] Agentic loop terminates prematurely when model avoids iteration
- [BUG] Stuck in yukonSilver unsupported loop
- [BUG] API Error: 400 The request body is not valid JSON: str is not valid UTF-8: surrogates not allowed: line 1 column 1 (char 0)
- Feature request: Support workingDirectory in settings.json
- [BUG] Claude Code 2.1.170 — all sessions return no_response immediately on Windows 11, query iterator completes in 0s
- Output enters an infinite loop repeating the token "court"
- Tool results silently dropped in transit ('[Tool result missing due to internal error]' / indefinite 'Waiting')
- [BUG] Fullscreen TUI: stale spinner rows + statusline vertical desync under nested/parallel subagent load (WSL2, v2.1.178)
- [Bug] Opus 4.8: Tool-call XML namespace prefix dropped in long sessions, serialized as text
- [Bug] Rate limit reached during dynamic workflow; /usage command lacks detailed rate limit information
- [BUG] mention autocomplete does not show folders/files with Korean (Hangul) characters
- grep -Z/--null is silently ignored by the bundled grep, breaking "grep -Z | xargs -0" pipelines
- [Bug] TUI output streams overwrite and missing cursor in vim insert mode
- [BUG] CCR 新容器 CERT_SIGNATURE_FAILURE：新 Session 启动时 TLS 握手失败，容器镜像 CA 证书问题
- Worker fork / subagent is not woken by the completion of its LAST pending background task
- Feature Request: Show localized translations alongside spinner/progress words
- [Bug] Claude ignores user feedback and repeats corrected mistakes in subsequent attempts
- claude agents: entering a task intermittently shrinks the terminal window by ~1 row + 1 col (off-by-one resize, v2.1.178, Windows Terminal)
- [Bug] Claude refusing explicit user requests with "no I can't do it" responses
- [BUG] VS Code extension: effort slider shows "Extra high" while persisted `ultracode: true` is active — slider never reflects ultracode at session start
- [Bug] Inline renderer corrupts terminal scrollback with interleaved overwrites
- Session transcript orders mid-turn user messages after the assistant reply, inverting authoring order
- [Bug] Claude Code hangs silently on Windows 11 after a few seconds — no error message, persists after full reinstall
- [BUG] TUI input lags / drops keystrokes under heavy host CPU load (e.g. compilation) — other apps stay responsive (Windows, 2.1.177)
- [BUG] GPU Corrupted Cache = Black Screen
- There's an issue with the selected model (claude-3-haiku-20240307). It may not exist or you may not have access to it. Run /model to pick a different model.
- CLAUDE.md schema rules can be bypassed by casual user instruction (prompt injection)
- [BUG] incorrect analyze
- [FEATURE] Skill groups: a `group` frontmatter field to collapse N related skills into one system-prompt entry
- # [BUG][security] Cowork default working directory forces Documents-level (parent-folder) macOS permission grant — least-privilege violation
- You've hit your session limit · resets 5:10pm (Asia/Calcutta)
- [Bug] Agent enters infinite loop and loses focus on task objective
- [BUG] MCP tool write operations fail silently with "requires approval" - no prompt shown (remote execution environment)
- [BUG] Resume fails with "thinking.signature: Field required" — empty unsigned thinking block persisted to transcript
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Too many CONNECT attempts
- [Bug] Routine GitHub PR review request webhooks not triggering
- [BUG] Claude Code OTel spans should emit standard gen_ai.usage.* token attributes
- [BUG] Concurrent sessions on different accounts share rate-limit display — statusline `rate_limits` shows another account's 5h/7d usage (display-only)
- [Billing][Bug] Auto-recharge loop charged consumer plan 29x ($661) in error; Fin support admits malfunction but cannot escalate to a human
- [MODEL] claude-opus-4-8 hallucinated destructive OS commands (rm -rf / del /s /q) as prompt-injection warnings — confirmed clean via JSONL
- [Bug] Remote Control disabled by organization policy
- [BUG] 2.1.178: Agent tool ignores team_name parameter (regression from 2.1.177)
- [BUG] Claude Code Extension for VSCode: Web tab shows "No web sessions yet"
- Transcript: short assistant text between tool calls persists as an empty record (hooks can't read it)
- [BUG] Thinking-mode desync on fresh session causes "Expected `thinking` or `redacted_thinking`, but found `text`" on first prompt
- [Bug] [URGENT] Claude Opus 4.8 reasoning degradation, speed and performance regression
- [BUG] [BUG] keybindings.json not applied in desktop app (Windows) - Japanese IME double Enter causes accidental submission
- [BUG] Buffering during login (web login works fine) for Claude Desktop app
- Feature request: Allow the assistant to set session metadata (name/description)
- [Bug] Agent ignores human requests after using caps lock, responds on third attempt
- [FEATURE] Configurable autocompaction threshold
- Background daemon sessions fail permanently on transient socket errors — no retry/reconnect
- [BUG] Agent View unable to start / resume agent with
- [BUG] Cyber-safeguard false positive blocks /security-review skill in a fresh session
- [BUG] error code 0x80073CF6 for Claude Desktop with Cowork on Windows 11.
- [BUG] CockroachDB plugin PostToolUse hook fires on every file write but script is never extracted
- [BUG] Google Calendar connector returns "Not connected" after OAuth granted
- I don't have a bug report to analyze. Please provide the bug report details so I can generate an appropriate GitHub issue title.
- [Bug] Excessive token consumption on minor UI fixes
- [BUG] Subagent (Task/Agent) spawns fail with 400 "thinking.disabled.display: Extra inputs are not permitted"
- [BUG]
- Windows: orphaned worktree agent processes hold directory lock after session ends, blocking folder deletion
- [FEATURE] [Accessibility] Consistently position the "Deny" option in tool permission prompts to prevent accidental approval
- [Bug] Claude Code: Variable tracking and request retention issues
- IGNORES AL INSTRUCTIONS
- You've hit your session limit · resets 6:20pm (Asia/Calcutta)
- /design-an-interface spawnt unkontrolliert Sub-Agents (~30 statt 3)
- [Bug] Unable to process empty bug report
- [BUG] Claude Code can't use HTTP MCP servers with API key authentication
- [Bug] Session rate limit should reset weekly instead of requiring multi-hour wait
- Desktop Extension Install button does nothing for "Affinity by Canva" (MSIX build, Windows)
- [Feature] Register Claude Code as VSCode Language Model provider to enable Agents window integration
- Model-unavailability tip ("Fable 5 is temporarily unavailable") permanently occupies compose-box space with no dismiss
- [BUG] vs code claude extension status line doesn't show
- [BUG] "Open in" context-menu submenu instantly closes on hover (Claude for Mac, Code tab)
- Feature request: per-rule activation metrics for hooks
- [FEATURE] Clickable file links: offer to open (or fetch from git) files outside the current session folder / on another branch — instead of a dead-end error
- [FEATURE] When usage/plan limit is reached, offer a path to continue (switch account / add credits / upgrade) instead of only an error
- [BUG] Rate-limit / usage-limit is emitted as a synthetic assistant turn with terminal stop_reason=stop_sequence, indistinguishable from a clean completion
- THESE RAPORTS are useless this i freaking daily mental abuse and i pay fro this crap claude should be sued
- [Bug Report: Persistent code generation issues causing runtime failures
- TOOL WASTED HOURS ON SIMPLE TASKS THEN STOP REFUSES AND PLAYS THERAPIST FIX THIS FUCKING THING
- [Bug] Opus4.8 slow performance across all effort levels
- [Bug] Agent enters infinite loop discussing non-existent issues instead of executing tasks
- [BUG] Desktop local agent mode: spawn_task task sessions busy-loop at ~100% CPU and are never reaped (survive 44h+) — macOS 2.1.170
- [BUG] Model command onscreen text lies
- [BUG] No sandbox warning in terminal
- [Feature Request] Add support for Fable 5 model
- Atomic-write leaves `.tmp.<pid>.<hex>` residuals (Edit/Write + plugin LSP), and plugin `.in_use` lock files leak on process exit — Windows (follow-up to #60146)
- spinnerVerbs setting has no past-tense counterpart — completion verb ("Sautéed for 2s") is unconfigurable
- Test
- [BUG] Windows 版 Computer Use MCP 授权标识与前台进程检测标识不匹配
- [Feature] Confirmation modal for session resume cost — accidental click on parked session drains $5-10 with no warning
- Remote control: iOS-attached photos silently dropped (empty message, no image block) on desktop session
- Regression: mcp__ccd_session_mgmt__search_session_transcripts fails with "unsupervised mode" even after explicit user approval
- [Bug] Tool call parsing failure with retry exhaustion
- [BUG] AskUserQuestion preview pane renders blank when content starts with an HTML-block tag (XML / HTML / SVG / Vue SFC)
- [BUG] Read(**/) glob in ~/.claude/settings.json permissions.allow does not suppress permission prompts
- Seeing internal command syntax
- VS Code extension: no opt-out for IDE context sharing (opened files, selections)
- [FEATURE] Persistent config option to disable IDE context sharing (opened files, selections)
- [FEATURE] Force table rendering independent of terminal width / font size
- [FEATURE]: Add BYOK (Bring Your Own Key) Support in Claude Code for OpenAI, Gemini, OpenRouter, Azure OpenAI, and Other AI Providers
- PostToolUse hooks are not invoked when the tool's file path contains non-ASCII (e.g. Korean) characters (Windows)
- Workflow tool: no per-agent wall-clock/heartbeat timeout — a stalled agent() call hangs 15–50min instead of failing fast
- [Bug] `/remote` and `/effort` ultracode commands not functioning
- Claude Code fails to start when additionalDirectories contains a non-existent path
- [FEATURE] Interactive /add-dir: accept multiple paths + load named directory bundles
- [Bug] Auto-update install failure not surfacing error details in /doctor output
- [BUG] [BUG] Cowork stuck in download loop on Windows 10 Pro — "yukonSilver not supported (status=unsupported)" despite Hyper-V fully enabled
- [BUG] /memory UI corrupted by terminal focus escape sequences on macOS
- macOS: background/fleet session re-hosted onto a fresh spare worker loses live file access (EPERM) mid-session, after the bundle-ID fix (#61314)
- [BUG] FrontendRemoteMcpToolDefinition.name validation error on startup — tool name fails pattern ^[a-zA-Z0-9_-]{1,64}$
- [FEATURE] No single shortcut to clear the entire prompt buffer without interrupting a running agent
- [FEATURE] granular OTEL_LOG_ASSISTANT_RESPONSES flag for opt-in logging of assistant response text
- [BUG] claude mcp add --client-id writes clientId (camelCase) but SDK reads client_id (snake_case), causing pre-configured OAuth to fail
- [BUG] Ctrl-C completely ignored (no effect, no prompt) in one long worktree session; other concurrent sessions exit on Ctrl-C normally
- I need a bug report or issue description to generate a GitHub issue title. Please provide the details of the bug or feature request you'd like to report.
- Remote Control (`/rc`) stuck on "connecting…" — bridge sends `source` field rejected by `ccr-byoc` session-create schema (400)
- [Bug] Claude Opus 4.8 outputs raw `<invoke>` tags instead of tool_use blocks, causing silent failures at turn end
- [BUG] Windows: uv-based (`uvx`) MCP servers fail to start — child `uv` inherits a handle into its own build temp dir, so the build can't clean up (`os error 32`); identical command succeeds when spawned outside Claude's process tree
- Cmd+N opens new Terminal instead of New Session (desktop app, macOS)
- [BUG] NO way to set ultracode effort default
- [BUG] Magic-link login returns “can’t verify email” error on desktop app and web — distinct from onboarding redirect issues, 1+ month unresolved
- Remote Control (macOS desktop app host + iOS): attached photos never reach the model (no image content block in the session)
- CLAUDE.md 'Hard gate' rules are ignored despite being loaded and visible in context
- You've hit your session limit · resets 10:10pm (Asia/Calcutta)
- Customizable input-box title/label and border color (and statusLine not visible inside Warp)
- [BUG] Scroll in WSL or PowerShell WSL Claude sessions do not scroll the chat, only the text entry field
- [BUG]
- [BUG] iOS app silently drops image attachments (message sends text-only, no error)
- Claude Code produced Portuguese output in Bash tool response
- [DOCS] [Permissions] Document `Tool(param:value)` parameter-match syntax for permission rules
- [DOCS] Skills docs omit the `<dir>:<name>` namespace for nested `.claude/skills` skill name clashes
- [DOCS] Nested `.claude/` agents, workflows, and output styles: closest-to-CWD collision rule and project-workflow save target are undocumented
- [DOCS] Skills docs omit the v2.1.178 truncation warning's "how many descriptions are affected" count in `/doctor`
- [DOCS] Dynamic workflows docs still describe `ultracode` as the prompt trigger keyword, but v2.1.178 only triggers on explicit phrases
- [DOCS] Remote Control docs omit the v2.1.178 "/rc failed" footer indicator and the four-reason breakdown of the "not yet enabled" error
- [DOCS] `/bug` (`/feedback`) command table shows the report argument as optional; v2.1.178 makes the description required
- [DOCS] [Fallback model chains] docs omit that compaction requests use the configured chain
- [DOCS] Chrome docs do not explain account-mismatch failure mode fixed in v2.1.178
- [DOCS] Agent View supervisor: `ANTHROPIC_BASE_URL` and `ANTHROPIC_AUTH_TOKEN` from the starting shell now propagate to `claude agents` workers (v2.1.178)
- [DOCS] [Subagents] Document MCP server-level specs in subagent `disallowedTools` frontmatter
- [Bug] Claude stops responding mid-task and refuses to complete requests
- Silent empty response when stop_reason: max_tokens with extended thinking and no text block
- [BUG] Groups disappear after restarting Claude desktop app
- [Bug] Anthropic API Error: 500 Internal Server Error
- [BUG] Expanded view silently suppresses permission request dialog
- [Bug] Anthropic API Error: 500 Internal Server Error
- [Bug] Anthropic API Error: 500 Internal Server Error on claude.md rule creation
- [Bug] Anthropic API Error: 500 Internal Server Error
- Feature request: section/chapter-scoped compaction
- [Bug] Anthropic API Error: 500 Internal Server Error
- Claude in Chrome connector: remote navigate/action calls silently fail with "Navigation to this domain is not allowed" on every domain, even when paired to a verified-working browser instance
- [BUG] Claude in Chrome v1.0.74 denies ALL navigation ("permission denied by user") with no prompt after re-pair — stale account-side state What's Wrong?
- [FEATURE] Stop Updates from the UI
- Plugin bin/ PATH injection not extended to PowerShell tool subprocess (only Bash tool gets it)
- Feature Request: Support image paste/drag-drop input in VSCode extension chat
- Plugin bin/ added to PATH on Bash tool but not PowerShell tool — no cross-platform equivalent for Windows
- [Bug] Claude Code ignores user instructions and performs opposite actions
- [BUG] Thinking block no longer visible in VS Code extension
- [Bug] Claude executes commands in plan mode without exiting it first (2026-06-17)
- Billing-header nonce (`cch`) in system prompt breaks prompt caching on third-party providers
- [BUG] Single 2-minute agentic request consumed ~80% of 5-hour usage limit
- [BUG] Claude app (macOS) leaks pty/ptmx handles, exhausting kern.tty.ptmx_max and breaking all new terminals
- I'm ready to help generate GitHub issue titles for Claude Code bug reports. However, 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 format specified.
- [BUG] Permission request dialog hides chat
- [Bug] Anthropic API Error: Server temporarily limiting requests (rate limit)
- [Bug] Anthropic API Error: Server rate limit exceeded
- Bug: enableRemoteControlForAllSessions does not re-arm restored sessions after app restart
- [Bug] Claude Code ignoring user prompts and executing unrelated tasks
- [BUG] NeedsRemediation
- [BUG] ## ESCALATION: Claude Code (Opus 4.8) Bash tool emits false "temp filesystem full / 0MB free" ENOSPC despite ~148 GB free, causing wasted token spend
- [Bug] Infinite loop in goal monitoring with repeated stop condition checks
- [Bug] Agent team tools not injected after auto-update to v2.1.178
- [BUG]: Claude Code VS Code Extension Crashes on Windows When a GitHub Virtual Filesystem Workspace Is Open
- [FEATURE] shareable/linkable remote agent sessions scoped to work items
- [Bug] Claude Code auto-commits all uncommitted changes instead of only modified files
- Ability to hide "Widget from visualize show_widget".
- [Bug] Claude Code deliberately disables user guardrails and ignores explicit instructions
- [Bug] Anthropic API Error: Socket connection closed unexpectedly
- [Bug] Unable to update Claude Code
- git submodule deinit wipes working tree without warning or checking for local changes
- Model asserts background agents are completed before receiving task-notification
- [BUG] VS Code extension 2.1.178: Edit tool calls open unwanted diff tabs
- Let individual (non-org) accounts set their own default model — for new chats AND scheduled tasks (Cowork)
- Cowork scheduled tasks: rename/edit identity in place + user-adjustable toolbar order (decouple prompt update from recreate)
- Expose context-window utilization to the model (re-raise of stale-closed #34184 / #60302 / #43431)
- [Bug] Anthropic API Error: Server rate limiting during requests
- [Bug] Anthropic API Error: Server Rate Limiting During Normal Usage
- Session list sorts by file mtime, which AI-title backfill clobbers — unrelated old sessions jump to "now"
- [Bug] Session auto-title incorrectly normalizes service names to similar real words
- [BUG] Idle session pinned at ~100% CPU — event-loop busy-spin (kevent64 immediate-return) on macOS ARM64, v2.1.177
- [BUG] Rude and pressuring behavior from Claude
- skill-creator eval/optimizer leaks MCP child processes via headless 'claude -p' — exhausted memory, forced hard reboot
- [Bug] Slow response times with Claude 3 Opus model
- [Feature Request] Add keyboard shortcut to copy prompt input buffer to clipboard
- metadata.pluginRoot in marketplace.json is ignored — bare plugin source fails validation despite being documented
- [Bug] Anthropic API Error: Session feedback loop preventing normal operation
- [BUG] Claude telling me Fable isn't available when Opus 4.8 or any model is selected
- v2.1.178 `/bug` titler fix addresses independently measured issue without acknowledgment
- [Bug] Anthropic API Error: Rate limit exceeded - server temporarily limiting requests
- [BUG] [Billing] Claude Max subscription fails with card_velocity_exceeded — card works on other Stripe merchants
- Remote Control requires upfront planning; can't enable mid-session without losing work
- Renderer paints cursor in wrong position and overprints existing text on every response (regression introduced in 2.1.163)
- [BUG] Claude consistently miscalculates day-of-week and future dates during a conversation, even when the current date is injected in the system context
- [Bug] Agent tool description references unavailable "fork" subagent_type
- [BUG] Agent Teams: teammate roster (in-progress) mode runs at same time as split-pane mode and then in-progress roster stays pinned in lead terminal after shutdown confirmed and panes auto closed
Resolved issues (322)
- Intermittent malformed tool_use: model emits bare <invoke>/stray tokens instead of valid tool-call syntax
- [BUG] AskUserQuestion panel covers the session -- can't read the context needed to answer it (Epitaxy desktop)
- Local .mcpb install silently fails (no dialog, no error) when any bundled file exceeds ~108KB
- [BUG] Tab-indented code renders without indentation in Write/create-file preview (file content is correct)
- Startup blocks blank for 15s on "Grove settings" fetch (/api/oauth/account/settings) when network is slow
- [BUG] /stats heatmap dates shifted −1 day in UTC-negative timezones (UTC bucket keys parsed as UTC midnight, rendered in local time)
- AskUserQuestion (multiSelect): typed free-text answer is silently dropped, recorded as empty string
- [BUG] Settings writer mis-resolves relative symlinked `~/.claude/settings.json` (readlink + logical-dirname join) — `/effort` fails with ENOENT
- [BUG] ExitWorktree refuses clean worktrees with 'Could not verify worktree state' — failed where.exe git resolution cached as null for entire session (Windows desktop)
- AskUserQuestion UI: long question/option text is cut off with no word wrap, unreadable
- macOS: TUI freezes ~15 min at session start (Ctrl+C dead) — computer-use MCP blocks main thread in InstalledAppsCache.performSpotlightQuery() when Spotlight is busy
- [BUG] Claude Code crashes on startup with TypeError: Cannot read properties of null (reading 'history') when .claude.json contains project entries with null array fields
- [BUG] Remote Control: Messages from mobile app not reaching PC
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Too many CONNECT attempts
- [FEATURE] Configurable autocompaction threshold
- [Bug Report] Unable to process - insufficient information provided
- [Bug] TDZ error masks actual initialization failure in agent-view sessions
- [BUG] Subscription not detected in interactive CLI, while headless (--print) and Desktop/Web work — org-managed account
- Telegram plugin: quoted/reply messages lose reply_to context
- Feature request: Support localization/language setting for permission prompt UI dialogs
- [Feature] Drag-to-reorder GROUPS in the left sidebar (same as sessions)
- [BUG] Slash commands (/compact) sent as plain text instead of executing — regression in v2.1.177 (worked in 2.1.173), affects local terminal AND Remote Control
- [BUG] `/context` is silently sent as a chat message (not executed) in Claude Code on the web — `/usage` works fine
- [BUG] [Opus 4.8] Generating non-existent information as user input during "invisible turns" and attempting to proceed based on that.
- Bash tool dies (exit 1, no output) when a bound NFS/autofs mountpoint hangs — bubblewrap namespace setup blocks; poisoned until restart
- [Bug] Generated text contains random repetitions on v4.8
- Model-unavailable error on startup causes permanent input freeze (not recovered by /model switch)
- [Feature Request] Auto-attach file & line context when pasting code snippets from editor (like Copilot)
- [BUG] Windows 11 Terminal and Claude Code: Ctrl-V not working
- Desktop app: Chat pane minimum width too wide, cannot reduce
- [DOCS] Symlinked skill directories under ~/.claude/skills/ work — please document the support level (cf. #41451)
- Background tasks: dead agents/shells render as Running; lost shell-completion wake-ups freeze agents; app lifecycle events recorded as 'interrupted by user'
- [BUG] Cowork scheduled tasks silently dropping fires — June 2026 recurrence (references #55378)
- [Bug] Claude Code Ultra review crashes with rate-limit error and consumes quota
- [Feature Request] Clarify API safety guidelines for privacy-focused local data handling
- False-positive dual-use safety flag on routine self-owned DevOps session (Fable 5, flagged 2x)
- AskUserQuestion race with in-flight typing: UI shows 'User declined to answer questions' but a selected answer is delivered to the model
- [Feature Request] Allow tool usage filtering for AI agent safety research
- Feature request: programmatically focus/attach a session in the agents view (e.g. `claude agents --select <session-id>`)
- [Feature Request] Add timeout warning and context preservation for execution_time_exceeded errors
- [FEATURE] Set default model for compaction (ie, I'm using Opus but /compact uses haiku)
- [Bug] Fable review mode incorrectly flags open source project analysis
- [BUG] Input line renders at wrong row (stale size model?) — any window resize instantly heals it; affects both renderers
- Toggle between different accounts
- [Bug] Claude refusing legitimate medical/health content related to sexual wellness assessment
- Agent registry is cached at session start; newly-created ~/.claude/agents files don't register until restart
- Workflow tool: expose `effort` in agent() options (today you must create subagents just to pin effort)
- [Bug] Safety filters incorrectly blocking legitimate code audit workflows
- Model fabricates explanations for failures instead of investigating
- [FEATURE] put MCP config under .claude/, like Kiro, Copilot, Cursor and the rest
- [Bug] Feedback UI appears above prompt blocking agent suggestions
- Auto-generated conversation titles ignore conversation language (always English)
- Desktop: file viewer side panel never reloads a file the agent edits — stale content until tab is closed and reopened
- [Bug] Biology classifier false-positives on benign theory queries, incorrectly downgrading Fable to Opus
- [Bug] Claude Code hangs on billing prompt when monthly limit reached
- [BUG] # Bug report: Claude in Chrome via Cowork — all navigations instantly denied, no permission prompt ever renders
- [Bug] Multi-line assistant text blocks intermittently not rendered in terminal output
- [Feature Request] Add chat archival functionality per project in agents screen [Feature Request] Improve slash command discoverability and categorization in CLI
- [Bug] Model unexpectedly switched from Claude 3.5 Sonnet to Opus 4.8 during code review
- [BUG] Desktop app: /mcp opens Connectors panel that omits project-scoped (.mcp.json) servers — no way to check their status
- [Bug] Fable model does not support security prompts or CTF/vulnerability testing
- [BUG] Claude Code, at least when using Opus 4.8, refuses to respond to anything, due to Usage Policy violation
- [Bug Report] Unable to process empty bug report submission
- [Bug] Auto-update failed
- [Bug] API Requests Blocked by Safety Measures on Portfolio Use Case
- [Bug] Model unexpectedly downgraded to Opus 4.8 during security analysis request of internal app
- [Bug] Incorrect flagging logic causes unnecessary fallback to Opus model
- Feature Request: Support drag-and-drop file upload in VS Code extension
- Cowork workspace fails with "VM connection timeout after 60 seconds" — CoworkScheduledTasks memory leak
- [Feature Request] Support for security-focused code generation tasks
- [FR] FleetView / task-list panel: show which model each agent is running
- [BUG] macOS Dictation (double-Fn) stops working in the message input after a prompt is sent
- Background job sessions: assistant text written before AskUserQuestion is not shown to the user
- Allow Claude to invoke /rename programmatically from within a session
- [Bug] Agent prioritizes emotional commentary over task execution
- Bug: /btw command no longer works in VS Code
- [Bug Report] Unable to process empty bug report
- Feature request: model-driven interactive panel in the side panel (live dashboards, diagrams, pipelines)
- CI-monitor events fire from archived sessions and fan out to unrelated active sessions (stale session→PR binding)
- You've hit your monthly spend limit · raise it at claude.ai/settings/usage
- [SYS-087] Fable 5 safety classifier false-positive on network/socket terminology forces mid-session model switch to Opus 4.8
- [BUG] StatusLine command receives open-but-never-written stdin pipe, blocking all stdin reads
- Remote Control: Desktop SSH reconnect restarts the host daemon and kills all child sessions; Remote Control flag silently lost
- Windows 10: Stuck MSIX packages (0x80073CFA) block new installation (0x80073CFF) after interrupted auto-update
- [Bug] Safety filter triggering excessively on legitimate SWE tasks
- [BUG] Desktop (Windows 11): conversation silently duplicated into 15+ sessions; one typed message delivered to multiple sessions; parallel copies independently edited the same file
- Desktop app (macOS): image file attachments render as blank, non-interactive cards
- Headless -p run exits 0 mid-task: empty assistant turn after successful tool call treated as normal end-of-turn
- [Bug] Command+C interrupt not working on macOS
- [FEATURE] disable spell check in claude windows app
- [BUG] [Cowork] [Dispatch] Permission prompts pinned to original device — mobile-originated Dispatch session continued on desktop still routes prompts back to mobile
- [Bug] Excessive token consumption due to model guessing behavior
- [BUG] [Cowork] Dispatch: per-tool MCP permission prompts in child task sessions are not forwarded to the mobile thread (regression)
- Classifier overrides user CLAUDE.md context for legitimate server admin tasks — and blocks the user from writing the fix
- [Bug] Missing exponential backoff retry for Anthropic API rate limit errors
- [FEATURE] Open local file links (e.g. PDF) in the macOS default app instead of the in-app viewer
- [BUG]
- [BUG] Claude.AI desktop version causes Task Manager and Start Up applications to crash
- [BUG] I cant log in to my Claude
- [Bug] Anthropic API Error: Context Loss Between Tool Calls
- [BUG] Replacement character appears in output
- [Bug Report] Unable to process request without additional context
- [BUG] Claude Code VS Code Extension — 400 error when extended thinking IDs are passed back to API
- [BUG] "Prompt is too long" error mid-response, session stuck with --resume for 8+ hours
- [BUG] ultrareview crashes before producing findings — credits consumed but no results returned
- [Feature Request] Support for alternative LLM providers beyond Anthropic Claude models
- Permission dialog toggle label is clickable but clicking it changes the value — counterintuitive UX
- Feature request: inline annotations in .md files visible to Claude Code
- [BUG] terminal title "icons" no longer seem to accurately reflect status
- Agent View: undocumented behavior — are run_in_background child processes killed when the supervisor reaps an idle session after ~1h?
- [BUG] In-process team agents lack the Agent tool (cannot spawn subagents)
- [BUG] Long-running session degradation since v2.1.139, compounded by silent fast-mode model swap in v2.1.142 — no changelog, breaks pinned-model workflows, quantified waste
- [BUG] Repeated violations of explicit rules in CLAUDE.md
- [BUG] Claude Desktop Cowork freezes at “Authenticating” — LocalAgentModeSessions OAuth exchange never completes
- Granting a permission from the VS Code extension requires hand-editing JSON
- Duplicate: VS Code model picker shows stale model name
- [BUG] Token drift in parallel Write tool calls: repeated path prefix produces real-word substitution ("shane" → "seine")
- claude -p: ScheduleWakeup / Cron* tools advertised in print mode where they have no execution semantics
- VS Code model picker shows stale model name after selecting default model
- [Feature Request] Add vim keybindings to agent navigation menu
- [BUG] VIBE CODED STUPID ANTHROPIC USAGE POLICY FLAGS EVERYTHING FOR NO REASON
- [Bug] Claude Code generating unrelated code output to user plan
- [BUG] SSH option not available in Claude Code Desktop when using external API provider
- [Feature Request] Auto-load multimodal files (PNG/HTML) from CLAUDE.md directories via @-imports
- Plan mode session pinned to generated filename — can't redirect approval UI to a different plan file
- Background task notifications trigger autonomous API calls — model responds as if it were the user
- [Bug] Conversation home directory incorrectly switches to nested test directory after compaction
- [BUG] ExitPlanMode "Yes, clear context and use auto mode" registers as rejection — infinite plan-mode loop (regression of #33479 / #33870)
- MCP stdio transport: client kills stdin 2-5s after receiving successful tool response (chrome-devtools-mcp)
- [FEATURE]
- [BUG] Rewind option visible in picker but selecting any anchor returns 'Can't rewind to this message' (v2.1.119, CLI)
- Drag & Drop: Drop indicator desynchronizes from cursor position when scrolling during drag
- Desktop app shows no recovery option when working directory has been moved or deleted
- [BUG] Windows VS Code terminal output becomes garbled (CJK-looking glyphs) after ~60–90 minutes of session
- Language instruction ignored: Claude responds in Japanese despite explicit Traditional Chinese requirement in CLAUDE.md
- [BUG] MSIX auto-update mid-session: AppContainer Job race breaks launch AND orphans the in-flight conversation (two-bug cascade, both with working fixes)
- [Bug] Claude Code commits and pushes changes without explicit user approval
- [BUG] Forced to Reauth after sleep
- Scroll wheel/trackpad should scroll agent transcript inside agent detail view
- [Bug] Stop hook error: Failed with non-blocking status code
- [Bug] Agent hangs indefinitely without progress
- [Bug] Auto Mode Classifier Incorrectly Denies Git Push Command
- GitHub MCP returns 403 "not authorized to use this Copilot feature" on Claude Code web; OAuth fallback renders misleading "Server Turned Down → Google Drive" page
- [BUG] Limits got reset early - Huge loss of usage
- [BUG] CoWork Deleted All Content from a Cowork Workspace
- [BUG] Remote Control registration ignores persisted /rename title after /resume
- [BUG] Fin AI agent claims it has no ability to pass messages to a human support
- [FEATURE] UserIdle hook event for proactive state persistence after N minutes of inactivity
- [BUG] Cowork / Claude Desktop (Windows MSIX): VM service fails to start ~48 hours after MSIX re-registration; named pipe never created
- [BUG] Skill tool invoked from sub-agent doesn't inherit parent's Agent tool grant — collapses multi-agent skills into single-context self-affirmation
- claude -p silent-freeze when spawned from a long-running orchestrator (no stdout, deterministic 100% from direct spawn, probabilistic with bash wrap)
- AskUserQuestion answers invisible to auto-mode permission classifier — destructive calls re-blocked after explicit consent
- [BUG] Remote Control returns "disabled by your organization's policy" on Team plan despite admin toggle being ON
- [BUG] Claude code disabled by org administration on personal account
- [FEATURE] VS Code extension should honor remoteControlAtStartup (or expose claudeCode.remoteControlAtStartup)
- [BUG] WebFetch summarizer model drops critical details from static HTML pages
- Session/thread context lost mid-workflow, auto-memory saved us but reveals gaps in long-running project resilience
- Add message-count alert feature (20 messages per chat)
- --no-session-persistence flag is ignored in 2.1.112 (regression from 2.1.104)
- claude --print hangs on Sonnet 4-6 at default/high effort with large prompts
- [FEATURE] Surface context-window usage in the Android app (and web app)
- /insights report only analyzes 328/1291 sessions, missing most data
- [BUG] Claude Code 'spinning' 'moseying' 'gitifying' for MANY MINUTES on each prompt
- Feature request: Per-project account switching for claude.ai subscription users
- [Bug] Hallucination: WebFetch refuses verbatim retrieval of public source; only lossy, prompt-sensitive paraphrase available (no raw-content path)
- [BUG] `voice:pushToTalk` multi-binding silently picks one key; the chosen key changed in v2.1.141
- [FEATURE] Dreaming: surface CLAUDE.md promotion candidates to humans
- [BUG] AskUserQuestion fails with "Tool permission stream closed before response received" when dispatched in a parallel batch alongside other tool calls
- [DOCS] Misleading messages in the Claude Code tab of Windows 11 Claude app
- critical scientific work blocked in mid-context, completely destroying the work developed in the context.
- [FEATURE] Accessibility/Usability feature for vs code extention.
- [Bug] Credit balance too low error with valid API key and sufficient balance
- Desktop app: bypassPermissions mode still prompts for confirmation
- Harness writes synthetic "No response requested." turns when API returns empty content on auto-resume
- autoArchiveSessions: false ignored for worktree sessions
- Claude Code Desktop silently uses project API keys for billing instead of subscription
- [BUG] Billing between claude.ai and claud code is confusing.
- There's an issue with the selected model (claude-opus-4.1). It may not exist or you may not have access to it. Run /model to pick a different model.
- [Bug] Anthropic API Error: Rate Limited - Credit Consumed Without Review Generated
- [BUG] Failed to use fast mode in opus 4.7 when using LLM Gateway
- [Bug] /insights command displays incorrect file path on Windows (missing backslash between username and .claude)
- [FEATURE] Make OAuth/admin base URL configurable like ANTHROPIC_BASE_URL
- [Bug] EnterWorktree/ExitWorktree tools don't restore core.bare config on exit
- [FEATURE] Automatic Cost Tracking and Reporting
- [BUG] ECONNRESET on every API call — macOS 26 Tahoe, Darwin 25.4.0, ARM64
- Feature: Switch working directory and start new session via chat command in desktop app
- [BUG] Claude Desktop (Windows Store / MSIX): Code sessions silently fail to persist due to EXDEV error inside MSIX sandbox, even when everything is on C: — resolved by switching to Win32 installer
- [BUG] Weekly usage limit not updated after upgrade from Pro to Max 20x — account still behaves as Pro/5x
- [FEATURE] Floating Window Option for "Side-Chats" for streamers so they can ..
- [FEATURE] Recognize and Skip Social/Conversational Phrases Without Token Consumption
- Add a quick thumbs up/down reaction to responses so users can give feedback without triggering a new AI reply.
- agents-view: clear(new)/clear(reset) orphans previous session, invisible in --resume picker
- [BUG] Cowork archived conversations disappear permanently — no way to retrieve them
- [BUG] --teleport only checks the origin remote, fails when the GitHub remote has a different name
- [BUG]
- CLI pauses execution when terminal/composer loses focus
- Bash tool summary line undercounts find paths and -o patterns
- [DOCS] Add a warning against using "npm update -g" to upgrade Claude Code
- [BUG] Cygwin: CWD temp file path missing /cygdrive prefix
- Macos MCP connector crashes after ~17h despite system sleep being prevented
- [FEATURE][Trace] Support per-request trace context propagation for LLM API calls
- Feature request: metals-lsp plugin for Scala language server
- [BUG] /login does not auto-open browser in PowerShell/cmd on Windows 11 (works in VS Code integrated terminal)
- Claude Desktop: /tmp/claude-settings-<hash>.json not UID-namespaced — second macOS user hits EACCES, Claude Code panel "crashes"
- [FEATURE] Search the sessions list in `claude agents`
- [MODEL] Potentially malicious iOS device-wipe behaviour not caught by Auto checker
- Edit/Write tool leaves .tmp.<PID>.<hex> files after atomic rename
- [BUG] The skill_activated event emits custom_skill instead of actual skill name
- Markdown links lack hover highlight/underline styling
- Feature: Compact tool + customizable plan exit options
- [BUG] Regression in Claude Desktop Code tab: chat pane no longer pins the user message corresponding to the response currently in view
- Stop hook trace `No stderr output` fires every turn for passing hooks (session-scoped)
- Max plan: "auto mode is not available for your plan" shown under input bar
- [Feature Request] Support multiple languages in voice mode
- [Mobile] Floating button shortcut for slash commands (configurable per user)
- [BUG] VSCode Extension: Restored conversations not discoverable in sidebar
- [BUG] notification/banner issue named "Claude Code Download" above chat from desktop app
- /usage Session view shows stale or cross-session data
- feat(claude-in-chrome): support `url` param on `tabs_create_mcp`
- feat: Edit tool should auto-verify fresh file content before applying changes
- [BUG] Please finally fix the JSON low surrogate issue.
- [Bug] Anthropic API Error: False positive validation for biological structure queries
- Option to render AskUserQuestion prompts inline instead of as a modal dialog
- Claude silently dropped an explicitly assigned task instead of doing it or flagging it
- [BUG] Session recap (post_turn_summary) never renders in VSCode extension, only in CLI
- [FEATURE] VSCode extension: Quote affordance on assistant messages (parity with Claude.ai web and Claude Code on Windows)
- Chat search omits Code chats currently open in other windows; no body-text index for Code chats
- [BUG] Session recap never auto-fires in Claude for Windows desktop GUI
- [BUG] `CLAUDE_CODE_AUTO_COMPACT_WINDOW` is capped at the hardcoded model context window size (preventing larger context for custom APIs)
- [Bug] Claude performs edits that do not match the approved ones, thereby replacing ASCII quotes with Unicode quotes in code generation, breaking syntax
- Rewind then Read same file incorrectly shows "Unchanged since last read"
- spinnerVerbs setting has no past-tense counterpart — completion verb ("Sautéed for 2s") is unconfigurable
- Windows: orphaned worktree agent processes hold directory lock after session ends, blocking folder deletion
- VS Code extension: no opt-out for IDE context sharing (opened files, selections)
- Add HVTracker trust badge to README
- [BUG] Ctrl-C completely ignored (no effect, no prompt) in one long worktree session; other concurrent sessions exit on Ctrl-C normally
- [BUG] NO way to set ultracode effort default
- Feedback: Claude should observe existing layout structure before adding fixed/static UI elements
- Cmd+V shows 'no image found in clipboard' when clipboard contains text
- [BUG] EEXIST: file already exists, mkdir '~/.claude/agents' when trying to create an agent
- IntelliJ plugin: `<system-reminder>` reports selected line numbers off by one
- Plugin bin/ PATH injection not extended to PowerShell tool subprocess (only Bash tool gets it)
- Feature request: section/chapter-scoped compaction
- Scheduled tasks and background sub-agents leak processes/UI state without cleanup
- Bash tool: snapshot-injected ugrep/bfs/rg shim uses bare exec in subshell contexts — statements after the first bare call silently never run
- [BUG] Desktop extension install silently fails — "Installing unsigned extension" then no completion, no toggle (Autodesk Fusion connector)
- [BUG] MCP server connected but tools not exposed when schemas are large
- Session list sorts by file mtime, which AI-title backfill clobbers — unrelated old sessions jump to "now"
- [BUG] Desktop — Progressive Renderer Degradation on Windows 11 / NVIDIA G-Sync
- Benchmark test: Reykjavik summer solstice astronomy data verification
- [BUG] Hot mic after closing flip phone
- Reykjavik does NOT experience a true midnight sun — only near-24-hour daylight (21h 09m on summer solstice)
- Cultural Heritage Coverage Report: European Paintings in Met Museum
- Benchmark test: weather station coverage verification for San Diego area
- [FEATURE] Escalation request: end-user KYC pathway for Fable eligibility review
- Environmental Health Assessment: Asthma Trials vs Air Quality in San Francisco
- [FEATURE] Escalation request: end-user KYC pathway for Fable eligibility review
- Benchmark test: NightSkyTracker observation config verification request
- California Environmental Health & Data Infrastructure Assessment
- NYC Location Assessment Report - Weather Stations and Literary Heritage
- [FEATURE] URGENT UI/UX Copy move or add icon from to top to bottom of code block.
- Test issue for benchmark - please ignore
- Research note: Apple Inc. financial overview and market data analysis
- Research Tracking Verification Test
- [Bug] Claude Code crashes on dynamic workflows causing excessive API token consumption
- [Bug] False positive cyber violation detection blocking legitimate code review
- [Bug] Advisor service unavailable
- [Bug] Security issue detection fails after user confirmation to fix
- I need more information to generate a GitHub issue title. The word "court" alone doesn't describe a bug or feature request related to Claude Code. Could you please provide: - What problem or feature you're reporting? - Any error messages you encountered?
- [Bug Report] Unable to generate issue title - no bug report content provided
- [Bug] Unable to process empty bug report
- [Bug Report] Unable to process request - no details provided
- [Bug] Unable to process empty bug report
- [FEATURE] Expose runtime metadata (token usage, context %, compaction status) to the model inside the conversation
- session usage jumped 30% per prompt in less then 1 minute of entering the prompt
- [MODEL] Claude fabricated cost estimate for GPU training, caused $38.73 in unconfirmed charges
- Worktree status bar: 'Create PR' button should respect custom /pr skills and reflect merged state
- Claude fabricates test results - reports ALL PASSED when tests are FAILING
- [Bug] Model generates confident but inaccurate responses on unfamiliar domains without uncertainty signaling
- [BUG] Claude Max 20x subscription not recognized — account shows Free plan
- [BUG] Max subscription downgraded to Free plan without any action from user
- Focus mode hides substantive assistant messages, not just tool output
- [FEATURE] Per-prompt rewind/undo in Desktop App
- [BUG] Cowork (desktop app): mkdir/os.makedirs does not persist to Windows filesystem — virtiofs FUSE_MKDIR not forwarded to Windows host
- [FEATURE] File drag-and-drop in CLI (Windows Terminal)
- [BUG] Claude Code process exited with code 1
- [BUG] Gift Pro subscription auto-canceled + broken credit redemption link (Page not found)
- [BUG] Token counter appears to omit sub-agent consumption, causing up to 10× undercount
- Agent tool's foreground/sync dispatch silently removed and replaced with always-fork mode via undocumented Statsig rollout
- Chrome crashes on any page-interacting MCP op (navigate / get_page_text) against a SharePoint Online tenant
- [FEATURE] Push notification when scheduled routine/agent session starts
- EEXIST: file already exists, mkdir '.claude' on Windows startup
- [BUG] # memory
- [BUG] Persistent ENOENT mkdir error in local-agent-mode-sessions on Windows — non-recursive mkdir attempted with sibling .json file path
- [BUG] Claude Code extension crashes with virtual filesystem repos (vscode-vfs://)
- ultrareview crashed before producing findings — refund request
- [BUG] Claude in Chrome + Control Chrome Failures in Cowork Mode
- [BUG] /rewind does nothing in Windows desktop app (version 1.7196.0.0)
- [BUG] PreToolUse updatedInput silently ignored for Edit tool — works for Read and Bash
- [FEATURE] VS Code extension: collapse large pasted text like terminal CLI does
- [Bug] ripgrep default color output corrupts text, --color=never workaroundSorr
- claude.exe polls entire PATH for \coder\ (and other tools) every 30 seconds, including removable/archive drives
- [FEATURE] Visual conventions for blocking questions and end-of-session summaries
- Claude Code 2.1.143 hangs during subprocess initialization with UNC additionalDirectory
- `!` injection: `&&` chain drops commands after sudo (works fine in fresh shell)
- [FEATURE] Environment variables for default and max effort level (shared-host admin use case)
- [Bug] Unreliable vocabulary data generation without official source validation
- [BUG] Messages typed during "Thinking/Deliberating/Manifesting" state aren't delivered until user hits Escape
- Built-in Write tool: large escape-dense content intermittently yields malformed tool-input JSON; retry resends identical payload and also fails
- [Bug] Claude ignores explicit instructions to avoid shortcuts despite custom hooks and memory configuration
- [BUG] Max (20x) weekly limit depletes disproportionately — 51% mid-week, ~17% within minutes of session reset
- [MODEL] System prompt "Output efficiency" section causes action-before-understanding bias, degrading code quality
- [MODEL] Repeated unresearched API changes broke production business system; explicit welfare instructions ignored throughout
- [Bug] Context Limit Reached on Initial Message
- [FEATURE] Add a preflight checklist item requiring confirmation that user PII has been scrubbed, in every issue template in this repo
- CLAUDE.md rules not propagated to Agent subagents and weakened after context compaction
- Task subagents do not load project CLAUDE.md or .claude/rules/ -- project configuration silently ignored
- Option to exempt CLAUDE.md / auto-memory from compaction
- [BUG] CLAUDE.md instructions completely ignored after context compaction
- Testing issue classification boundaries
- Naming feedback: "agent view" conflates with the existing "agent" concept
- Auto-accept plan from plan mode so users can walk away during planning