Digest / July 3, 2026
Claude Code Issues Digest — July 3, 2026
391 new issues filed ·
315 resolved in anthropics/claude-code (UTC day).
New issues (391)
- Peer-message security boilerplate fires on idle_notification status pings, spamming mobile UI
- [Bug] Safeguard activation triggered in non-agentic message context
- [Bug] Safety classifier false positive in code review workflow with Fable 5
- [Bug] AskUserQuestion auto-advances while user is actively typing, discarding input
- [Feature Request] Support standard OAuth/license authentication flows without safeguard triggers
- [Bug] Model automatically downgraded from Claude Fable to Opus without user request
- Launch preview panel auto-opens edited Angular component template .html files as static pages, rendering raw source
- [FEATURE] Expose sub-agent model in the subagentStatusLine payload
- [Bug] Anthropic API Error: Server temporarily limiting requests (rate limit)
- [Bug] Anthropic API Error: Server Rate Limiting - Temporary Request Limit
- [Bug] Anthropic API Error: Server Rate Limiting During Normal Usage
- [Bug] Anthropic API Error: Server rate limiting temporary request throttling
- [Bug] Anthropic API Error: Server temporarily rate limiting requests
- [Bug] Anthropic API Error: Server Rate Limiting - Temporary Request Limit
- [FEATURE] Process queued messages sequentially (one message = one turn, FIFO) instead of merging into a single combined message
- Desktop app: session status dot stays "running" after turn and all background tasks complete
- [BUG] Cowork MCP connector tools permanently disabled — only fix is renaming the connector key
- [Feature Request] Make AskUserQuestion timeout configurable
- You've hit your session limit · resets 2:50pm (Asia/Seoul)
- [Bug] Security vulnerability scanning restriction contradicts documented policy
- [BUG] Gift Max 20x subscription reverted to Free plan — $1,260 credit missing before gift period ended
- You've hit your session limit · resets 2:50pm (Asia/Seoul)
- Feature request: PreResponse hook to gate outbound assistant prose
- [BUG] Daemon supervisor shuts down for auto-upgrade but never restarts — bg workers orphan-watchdog-killed 60s later, all in-flight sessions lost (2.1.198→2.1.199)
- Windows: bang (!) prompt commands intermittently produce no output - silent failure, no error, self-recovered after ~1 day
- [Bug] Claude 3.5 Sonnet incorrectly escalated to Opus 4 on safe message
- [BUG] Desktop: personal git-marketplace plugins never auto-update despite autoUpdate:true; Update button is a silent no-op; stale gitCommitSha after CLI update
- /goal completion Stop hook re-matches phrases from earlier in the transcript forever, preventing clean session termination
- Ghost remote sessions persist in "Remote Control" panel with no way to dismiss them
- [BUG] Bash tool produces no stdout/stderr on Windows (working directory: C:\Projects\version)
- [Bug] Tool call failure causes Claude Code to stop functioning
- Routine status indicator (flashing dot/spinner) in sidebar is inaccurate
- [FEATURE][Desktop] Keyboard shortcut to move focus between split-view session panes
- [Bug] Request flagged without clear violation or error message
- [Feature Request] Auto-update plugins/skills at session start instead of on-demand trigger
- claude.ai MCP connectors (Google Calendar, Gmail, Drive) appear uninvited and nag every startup
- [Bug] IDE integration opens redundant integrated terminal host when inserting @-mention with external session active
- I don't see a bug report in your message. Please provide the bug report details so I can generate an appropriate GitHub issue title for Claude Code.
- [DOCS] `claude agents` over SSH on macOS fails to cold-start with "Could not switch to audit session" — troubleshooting entry missing in agent view docs
- [DOCS] `claude stop <id>` reliability against supervisor respawn is not documented — agent view docs describe supervisor restart behavior without stating that stops are honored
- [DOCS] `sub-agents.md` does not document the surplus-idle summary row collapse behavior in the subagent panel
- [DOCS] [Agent view] "From inside a session" backgrounding section omits that the session's `/color` carries over to the new row
- [DOCS] Startup recovery dialog behavior for corrupted config files is undocumented
- [DOCS] Agent view version history omits the v2.1.199 change to PR-link row rendering
- [Feature Request] Add configurable timeout or disable auto-continue behavior for user prompts
- [BUG] Possible bypass of Fable 5 guardrails
- [Feature Request] Improve Claude model demotion logic to consider context in security testing scenarios
- [Windows] AppX update/relaunch fails with "another program is using this file" (0x80073d02) — CoworkVMService (cowork-svc.exe) holds package file lock
- Cowork failure - Claude Pro
- [BUG] Windows: Read tool PDF fails with 'pdftoppm is not installed' only in VS Code extension sessions - same binary/env/cwd succeeds headless (-p); negative resolver result cached for process lifetime
- [BUG] Project-scope plugin installs are unremovable: uninstall says "enabled at project scope", disable says "already disabled" (still repro on v2.1.199; #62966 stale-closed)
- [Bug] Security safeguards incorrectly flag defensive vulnerability analysis as offensive content
- Feature Request: Cross-device memory sync for same user account
- [Feature Request] Implement token refund mechanism for model switch false alarms
- [BUG] Bash tool permanently broken with WSL2: unexpected EOF while looking for matching quote
- [Windows 2.1.199] VSCode extension spawns a leftover blank Chat channel after Reload/Restart (Root Cause: webview issues launch_claude before session-restore handshake resolves)
- [Bug] Git history cleanup triggers false positive on Fable security check
- [BUG] Model chosen via /model silently reverts to default when session continues from another device (quota still billed to pinned model)
- [Feature Request] Weather Radar Denoiser ML Model Design
- [Bug] Agent tool calls silently dropped in parallel <invoke> blocks
- [Bug] Anthropic API Error: Invalid file type or size limit exceeded for kernel screenshot/video analysis
- Claude Code (Opus 4.8) fabricates fake "injected instructions" and invents user rules, then refuses to work
- [BUG] Console window briefly flashes on every shell command execution (Windows)
- Subagent context % display uses 200k denominator for 1M-context ([1m]) models — pins at 100% while real usage is ~20-60%
- abrupt consumption of limits
- [Bug] Anthropic API Error: insufficient_scope not triggering step-up authorization flow
- [Bug] Model self-initiates ToS-violating detection evasion (TLS fingerprint spoofing) under benign task framing without consent
- [BUG] Windows Desktop app: Read tool PDF fails with 'pdftoppm is not installed' while poppler is installed, on PATH, and resolvable by the same session's shells
- [FEATURE] Claude Code session discovery – hypergraph-based navigation feature request:.
- [Feature Request] Unified autocomplete confirmation keybinding for skills and files
- Plan mode not enforced: Write/Edit/Bash mutations execute without prompts while plan-mode reminders active (2.1.183, cli entrypoint); spurious "Exited Plan Mode" events without user approval
- [BUG] Shell grep shim: BRE alternation \| combined with literal ( produces 'rg: regex parse error: unclosed group'
- Fullscreen TUI is repeatedly re-prompted and force-enabled on background session attach, overriding explicit user opt-out
- Feature request: render spinner tips (spinnerTipsOverride) in the VS Code extension panel, like spinnerVerbs
- [Bug] Model unexpectedly switches from Fable to Opus
- [Bug] Agent tool calling failed repeatedly
- SIGKILL from macOS 24h disk-write budget mislabeled as "blocked by endpoint security" on unmanaged personal Mac (~2.1 GB launch write burst, Cowork VM pre-warm)
- [Bug] Benign request incorrectly flagged as unsafe
- [BUG] Desktop app top-bar diff stat shows stale +N -0 after git reset --hard to remote (recurrence of #36364)
- statusLine command: terminal_width missing from JSON input
- [MODEL] Opus 4.8: fresh subagent's first turn opened with a self-fabricated "System:" prompt-injection payload, then refused it in the same completion
- [BUG] Transcript inspection of "ccd_session_mgmt: send message" message for deleted session is unexpected
- [Bug] Model unexpectedly switched from Fabel5 to Opus during code review request
- Workflow({scriptPath}) false-fails "hidden control characters" on a clean file with CRLF line endings
- Workflow({name}) does not discover saved workflows under .claude/workflows/, contradicting the tool description
- [Bug] Model switching fails when token quota exceeded
- {KLM infolinia polska 24h!@!}::Jak dodzwonić się do KLM?
- [BUG] Claude Desktop white/blank screen after update on Windows 10, MSIX AddPackage 0x80073CF6 during install
- [Bug] Model unexpectedly switched to Opus during multi-page HTML refactoring task
- Native UI (VSCode extension): no way to expand/view full thinking-block detail
- I don't see a bug report or technical issue in your message. Could you please provide details about the problem you're experiencing with Claude Code? For example: - What error message did you receive? - What command were you running? - What unexpected beh
- [Bug] Model reverts to Opus 4.8 when set to Fable 5
- [Bug] AutoMode Task Classifier Authorization Bypass: Blocked Operations Still Executed by Main Agent
- [Bug] Claude Fable 5 model reverts to Claude 3 Opus due to overly broad safeguard filtering
- [BUG] [Windows] Local Routines leak orphaned claude.exe processes over time — RAM grows unbounded and app reports phantom active sessions blocking clean shutdown
- Daemon-hosted session forks pin --effort from cached session state, ignoring settings.json changes
- You've hit your session limit · resets 3:30pm (Asia/Almaty)
- Agent Teams (tmux mode): named teammates start in $HOME instead of inheriting the lead's working directory
- [Bug] Anthropic API Safety Filter: Excessive Content Blocking
- [Bug] False positive safety filter flags legitimate tooling discussion as RE
- Feature request: bind keyboard shortcuts to custom slash commands in keybindings.json
- Feature request: user-configurable top-bar buttons to invoke custom commands/agents
- M365 MCP OAuth flow broken: duplicate 'prompt' parameter causes AADSTS900971
- [Bug] Terminal becomes unresponsive after viewing global memory file in VSCode
- [BUG] Claude Code blocks all models with "0% remaining" after 5 hour clock reset while /usage shows 3% session used / 97% remaining
- Desktop app v1.18286.0 launched in Spanish on an English-only macOS system
- [MODEL] Fable 5 is back - but completely unusable
- Attaching to a stopped/idle background agent crashes the worker with contradictory "currently running as a background agent" error
- [BUG] The prompt for routines created by API is not visible in UI
- [Bug] Nested subagents lost when switching to agents view and returning to session
- Fabricated <system-reminder>-style message falsely claimed files were reverted, instructed agent to conceal this from user
- Intermittent ECONNRESET / Connection error on /v1/messages — auth OK, transport-level reset, 3 macOS machines
- [BUG] Abnormal prompt-cache token growth across two consecutive turns in a single sessio
- Agents View shows wrong status (needs-input with no question, false red/failed, stale green) — overloaded state + stale exit-cause artifacts
- [BUG] Cowork: renaming a scheduled task doesn't persist — reverts to old name on restart
- Claude-in-Chrome: extension signed in with correct account, but MCP always reports "Browser extension is not connected"
- [Bug] Model reverts to Opus unexpectedly during session
- [Feature Request] Improve image manipulation capability for complex composition changes
- [Bug] Incorrect model warning: Opus warnings referencing Fable limits
- Terminal mouse reporting breaks cmd-click file links inside Claude Code TUI
- [Feature Request] Restore "Impeccable" skill weighting for web design polish argument
- [BUG] Desktop (Win): oversized main.log triggers rotation-check loop that never issues the rename — ~8-9% CPU at idle; log lines doubled
- [Bug] CUDA kernel troubleshooting issue
- statusline: expose per-model weekly rate limits (Opus/Sonnet/Fable) to the status line command
- [Bug Report: Fable compatibility issue with workflow requirements]
- [BUG] In auto mode, AskUserQuestion auto-selects the recommended option after ~60s of no user input (started in v2.1.198)
- [Billing Bug] Repeated unauthorized Max upgrade and charge — refund denied by Fin AI Agent citing "already refunded" policy
- [BUG] Skill tool re-invoked after slash command already loaded skill
- [BUG] Org-level banned flag not cleared after user-level reinstatement — cannot resubscribe to Max (account_banned on PUT /subscription)
- [BUG] Teleported remote-control session unresumable: every message fails with 400 diagnostics.previous_message_id
- [BUG] Remote control: background subagent work is invisible (no history, silent kill on session limit, stale View DIFF/PR chips)
- [Bug] File edit prompts appear despite shift-tab enabled and edit operations in allowlist
- [Bug] Fable 5 safeguard flag fires on every message in normal conversations, not "sometimes"
- [BUG] Claude Code process exits with code 3221225477 (0xC0000005) on Windows 11 25H2 during file operations, with "unsupported Windows version" dialog
- [BUG] Org-level banned flag not cleared after user-level reinstatement — cannot resubscribe to Max, blocks Claude Code CLI workflow (account_banned on PUT /subscription)
- [Cloud sessions] Regression between 2026-06-18 and 2026-06-25: GitHub REST passthrough returns 403 "GitHub access is not enabled for this session" for org repos despite Claude GitHub App installed
- [BUG] Org-level banned flag not cleared after user-level reinstatement — cannot resubscribe to Max, Claude Code CLI stuck on Free tier (account_banned)
- [BUG] Fable 5 safeguards repeatedly flag benign messages in a legitimate anti-fraud (T&S) session — forced fallback to Opus 4.8
- Windows: statusline cancellation (taskkill /T) and cross-session liveness probes overload WMI - WmiPrvSE pinned at 300-550% CPU with ~10 concurrent sessions
- [BUG] sandbox.enableWeakerNestedSandbox silently forced to false when CLAUDE_CODE_SUBPROCESS_ENV_SCRUB is set (breaks Bash sandbox in cap-dropped nested Docker)
- [Bug] Claude Desktop 1.18286.0 silently drops an entire MCP server if its manifest display_name contains "&"
- Garbled [Dynamic Context] block leaks shredded system-prompt/skill content into tool output after auto-compaction
- Add a compact/collapsed-by-default mode for tool call cards and thinking labels
- Android app: Code tab empty while claude.ai/code web shows all sessions (same account)
- DX: EnterWorktree nests worktrees inside the repo (.claude/worktrees/), causing editor/tooling friction
- Slack MCP plugin: Add conversations.mark (mark channel as read) support
- You've hit your session limit · resets 4:30pm (Africa/Nairobi)
- I need more information to generate a GitHub issue title. Please provide details about the bug or feature request you'd like to report, including: - What problem occurred? - What were you trying to do? - What error message (if any) did you see? - What beh
- $1,771 charged in 4 hours against a $25 spending limit
- [BUG] Project-scoped plugin enabled in multiple repos never loads on first session after switching repos (single install record, re-stamped after plugin load)
- [Feature Request] Add session persistence and automatic re-authentication across CLI invocations
- [Bug] Model configuration not persisting, reverting to fallback model
- [MODEL] Opus 4.8 confabulates a present-tense user request on turn 1 (cold start) — two sessions 13 min apart, JSONL-verified — plus a mitigation pattern that works
- Long sessions drift days past the session-start date — inject current timestamp per user prompt
- Feature idea: agent-mediated feedback with curated aggregation visible back to users
- Sync Claude Projects (planning) with Claude Code (execution)
- [BUG] False-positive Usage Policy block on a legitimate academic citation (surname collision) — also terminated an unrelated background job
- Enter key opens folder dialog instead of sending message in Code tab (Desktop App, Windows)
- Assistant inter-tool narration rendered in Japanese during a Traditional-Chinese-only (zh-Hant) session
- Schedules tab shows "No scheduled tasks yet" while scheduled tasks run correctly in the background
- [Bug] Musculoskeletal Simulation Task Failing
- [BUG] Tool calls repeatedly emitted with bare <invoke> tags (missing antml: namespace), causing parse failures
- [BUG] PyCharm IDE plugin: diff-accept no longer triggers automatic continuation (regression since recent update)
- [Feature Request] Remove time limits from interactive questions
- /plan in agent-overview mode silently discards typed input on error, with no recovery trail
- [Bug] Audio permission handling causes unexpected app context switching
- Option to disable interpreting image paths in pasted text as image content
- Telegram MCP plugin doesn't exit when Claude Code closes (orphaned bun server.ts processes accumulate over uptime)
- Add Arabic language support to /voice dictation
- [Bug] agent autonomously changed directory and began to write code unprompted and committed unprompted
- [BUG]
- Hier ist eine Description, direkt im Format der anderen.../
- Two separate billing accounts rate-limited simultaneously when on same IP — independent quotas not respected
- [Bug] Both weekly rate limit and Fable rate limit consumed while using Opus
- [Bug] API Error 400: role 'system' must precede an 'assistant' message
- [FEATURE] What is Anthropic doing? WE NEED A CLAUDE CODE KEYBOARD
- [Bug] Anthropic API Error: Context Limit Exceeded - Compaction Fails with Usage Credits Required
- [BUG] EnterWorktree fails for sub-repo worktrees in multi-repo workspaces (background session isolation deadlock)
- [Bug] Model switches from non-Opus models due to false positive safety flag
- Artifact tool: allow setting sharing/visibility programmatically
- [BUG] Claude Desktop 1.18286.0: tool-call cards render fully expanded and can no longer be collapsed (regression; bundled CC 2.1.197)
- [BUG] The Claude Desktop app (Linux beta) never displays the correct permission mode or effort level for sessions spawned from the Claude Code CLI
- Nested background agents recursively spawn sub-agents, loop on filler no-ops for 6.5+ hours, and become unreachable/unstoppable once the spawning agent's session ends
- [BUG] Incorrect version in update message
- Expose "export transcript to $EDITOR" as a bindable keybindings.json action
- AskUserQuestion: click meant to focus the window is captured as an answer selection (user doesn't know which option was submitted)
- [Bug] Fable 5 fallback to Opus 4.8 preserves effort level instead of mapping to capability-equivalent tier
- [BUG] Desktop app artifact-preview renderer balloons to 7.6 GB, driving the macOS memory pressure that makes the Claude Code daemon evict background sessions
- I need more information to create an accurate GitHub issue title. Your message appears incomplete or unclear. Could you please provide: 1. What is the specific bug or problem you're experiencing? 2. What error message (if any) are you seeing? 3. What wer
- [BUG] Unauthorized Pro→Max 20x upgrade, refund denied
- Markdown inline code (codespans) ignore custom theme overrides — make the color themeable
- [Bug] Safeguards flag triggering unexpectedly on CLAUDE.md files
- [FEATURE] Drag-and-drop folders into the prompt to attach full folder context (VS Code extension)
- [Feature Request] Improve content policy transparency and appeals process for enterprise users
- [BUG] Claude Code Desktop always shows Powershell tool calls as "Using Powershell" instead of showing provided description
- [FEATURE] Add optional character/avatar icons for Claude Code
- [Feature Request] Allow fuzz testing without guardrails in Fable
- [Feature Request] Add support for advanced code audit and security analysis with Fable 5
- [BUG] Cloud routines (remote_trigger) never load MCP connector tools — all connectors enabledInChat:false on 2.1.199 (regression of #35899)
- [Bug] Full-screen renderer loses track of long text entries and corrupts status area
- [Bug] Model safeguards triggered with insufficient reasoning provided
- [BUG] Inherited CLAUDE_CODE_CHILD_SESSION silently disables all transcript persistence - total conversation loss on process exit
- [FEATURE] Suporte nativo ao idioma Português no Claude Code — Direito Linguístico, Acessibilidade Universal e Inclusão Digital
- [BUG] Bug type: Segmentation fault (native crash, not a JS exception)
- LSP tool sends requests without checking server capabilities (issues methods the server never advertised)
- [BUG] Cowork: adding a folder to an ongoing session fails with "overlaps a protected host location", but creating a new workspace in the same folder works
- [VSCode] Drag-and-drop file reference inserts absolute path instead of workspace-relative path
- [Bug] Model auto-rerouting prevents task completion on simple code review requests
- [Bug] Search tool hallucinating results instead of executing query
- [Bug] Claude API returns malformed XML response starting with "court"
- [FEATURE]
- [BUG] claude-code in linux pycharm console send phantom command and duplicate copy-paste x11 function - i'm done with this bug !!!
- [BUG]
- [BUG] sessions/conversations dissapear and forking breaks in vscode plugin UX after using EnterWorktree tool
- [BUG] /login reports "Login successful" but credentials are never written to macOS Keychain (v2.1.199)
- [Bug] Claude repeatedly claims to have reviewed files without actually executing read operations
- [BUG] Cowork scheduled tasks fail when Slack is the first (or only) connector called
- [Feature Request] Add explicit settings toggle to disable timeout feature and document default behavior
- Custom subagent .claude/agents/tdd-worker.md never registers, identical siblings work fine (Agent type 'X' not found)
- Allow customizing spinner "thinking" verbs — static list and dynamic sources
- Custom Chat-context `enter` keybinding makes footer items unopenable (shadows Footer's default `enter` / footer:openSelected)
- I appreciate you reaching out, but this message is about a WiFi connection issue unrelated to Claude Code, the Anthropic CLI tool. If you're experiencing issues with Claude Code specifically, please provide: - What command you ran - The exact error messag
- [Bug] Anthropic API Error: Policy Violation on Legitimate Software Installation Request
- [Bug] Incorrect Fable 5 limit error when using Opus model
- [Bug] RAG Tool Duplication: Claude Recreates Previously Cataloged Tools
- [Bug] Anthropic API Error: Guardrails triggered on harmless research paper processing task
- [Bug] You've hit your session limit · resets 4:10pm (America/Sao_Paulo)
- [Bug] Excessive instruction repetition in task execution causing redundant "read the clone" operations
- [Bug] Security audit request flagged as inappropriate despite user ownership verification
- 투자도구 연결
- [BUG] GitHub connector shows authorized in 'Authorized GitHub Apps' but never appears in 'Installed GitHub Apps'
- [BUG] Mouse wheel scroll triggers prompt history navigation instead of scrolling conversation (macOS + Warp)
- [Bug] Subagent per-call model overrides not preserved on session resume; /status displays session model instead of actual subagent model
- Windows: local 'CLI bridge' stays cold after brief network blip; app restart does not recover it, only full reinstall/reboot
- Mid-session model switch not applied — UI shows the new model, session keeps running the previous one (+ silent budget loss)
- PowerShell safety guard false positive: here-string body text with paths like /requirements.txt blocked as 'Remove-Item on system path'
- claude-in-chrome: extension never registers with relay despite correct account and working sidebar
- [Bug] Safety filter false positive on defensive security code review terminology
- MCP config: HTTP server entry with `url` but no `type` is silently dropped — no warning, even under --strict-mcp-config
- [BUG] Custom fileSuggestion command: .map() passes the array index as each file's score, so MCP resources outrank every file except the first
- `/exit` hangs forever on "Cleaning up worktree…" when `claude --worktree` was launched from inside the worktree
- [Bug] Terminal clicks registering as input selection instead of focus action
- [BUG] VS Code Extension: Chat state does not sync between dedicated view and native VS Code chat panel
- Claude in Chrome extension not connecting (tabs_context_mcp fails)
- [Bug] AskUserQuestion tool auto-proceeds after 60s timeout without user confirmation
- [Bug] Safety guard triggered incorrectly without valid cause
- Agent tool subagents can access file paths outside approved permission boundaries
- [Feature Request] Claude refuses to assist with catless exhaust simulator development for legitimate trackday use
- [Bug] False positive safety flagging when working with DNA analysis code
- Fable safety heuristic false-positives on legitimate defensive own-fleet security engineering
- iOS/iPadOS app: assistant text in ScheduleWakeup-driven loop turns is hidden (collapsed into 'Ran a command, used a tool') — visible on desktop web + CLI
- [Feature Request] Add cybersecurity audit and vulnerability analysis capabilities
- Workflow tool: args object silently arrives as a JSON string inside the script, causing silent array/string type confusion
- archive_session("self") deletes the worktree but resumes the session onto it instead of terminating
- Windows: PowerShell/Bash tool-call windows pop up and steal focus, especially during scheduled/background task execution
- [Bug] Rate limit errors not handled gracefully
- Chrome integration: CLI never dials the native-host bridge socket — browser tools always report "Browser extension is not connected" (macOS)
- [Bug] Claude Code fails to execute with Fable 5 model
- [Bug] Unexpected Fallback Triggered During Model Usage Testing
- Bug: Cannot set claude-sonnet-5 as model
- Cowork: personal marketplace never re-syncs — "Update" reports latest version while repo is 3 versions ahead
- [Bug] Model refusing requests for legitimate security application
- [Bug] Claude Model Safeguard Incorrectly Blocks Legitimate Cryptography Operations
- NEW MODEL SAME LACK OF EXECTION
- [BUG] Skills named `agent-*` render with a stray `*` and wrong row style in the slash menu (id prefix collision with agent-mention suggestions)
- Session disappeared from sidebar + sidebar language switched from Japanese to English (Desktop app)
- [Bug] Model generating irrelevant responses and poor decision-making
- Marketplace plugin updates load newly-pulled executable code under initial-install trust, with no re-consent or diff
- [Bug] False positive completion status reported for active task
- Background sub-agents spawned by a subagent stay "Running" forever after the parent completes; TaskStop can't reach them
- [FEATURE] Only prompt workspace trust when configuration actually changes
- I don't see a bug report or technical issue here. This appears to be a general question about career advice for smart contract auditing, which isn't related to Claude Code (Anthropic's CLI tool). If you're experiencing a specific bug or issue with Claude
- Phantom user message (never typed, absent from session jsonl) appeared in model context
- [Bug] Claude Code executes tasks without explicit user permission despite configured mode settings
- [Bug] Claude Code executes tasks without explicit user permission despite configured mode settings
- [Bug] Model downgrade triggered on benign request without clear justification
- [Bug] Advisor returns "unavailable" deterministically when the request contains ToolSearch-loaded deferred tools — tiny context, both Fable 5 and Opus 4.8 executors (distinct from #67609's size threshold)
- [BUG] SessionEnd hooks never fire for long-lived `remote-control` sessions — lifecycle automation silently dead
- [BUG]
- --bare auth failure says "Please run /login" but /login cannot fix it — message should point to ANTHROPIC_API_KEY
- Desktop app scheduled tasks: missed-run catch-up dispatch silently dropped during app startup (no retry); lastRunAt stamped at dispatch; no 529 retry mid-run
- Desktop App: Allow choosing working directory before first message
- [BUG] Fable 5 = "Self-caged" : the model's own output (`fork`/`kill`/`orphan`/`reap`) trips its own classifier midstream — input rewording can't fix it
- [BUG] Session-to-project-group memberships silently dropped after app update; groups survive but sessions fall into Ungrouped
- [FEATURE] Per-subagent availability fallback model (declarative), not just silent inherit
- AskUserQuestion option chips intermittently fail to register a click
- [Bug] False positive self-flagging on blockchain project code
- [BUG] pasting a JPEG copied from Finder inserts file's image icon preview instead of the file
- Feature request: local historical per-model token usage (beyond current session)
- [Bug Report: Rule validation and execution issues in Claude Code CLI
- Account switch reverts active session .jsonl transcript and silently stops disk writes
- [BUG] Cowork crashes with exit code 1 on Windows 10 Pro - Claude Desktop 1.18286.0.0
- Desktop app renders ReportFindings (code-review results) as raw JSON instead of the findings panel
- automatic copy and paste issue
- [Feature Request] Allow Claude Code usage for authorized security audits and code reviews
- [FEATURE] Allow readline / emacs style line editing consistently
- VS Code extension: assistant message starting with `138.` renders as `38.`, dropping leading digit
- [BUG] Cursor flashes on every screen update (even while typing)
- [Bug] Overly restrictive safety flagging on legitimate ML model tuning tasks
- [MODEL] Sonnet 5 - ScheduleWakeup called repeatedly without ever ending the turn
- [Bug] Safety harness triggered during code review of generated features
- [FEATURE] Terminal client for Remote Control sessions — `claude connect <session>` to attach from the CLI like the mobile/desktop apps do
- API 400 error al pegar capturas tomadas con Alt+Impr Pant (Windows)
- Bash tool drops all stdout on Windows/Git Bash intermittently
- Write tool result for new files renders without added-line (green) highlighting, unlike Edit
- [BUG] Marketplace plugin re-materialization mid-session cleans the pinned hook installPath, silently killing hooks in every running session
- Fable 5 autonomous session destroyed user's main app via unverified assumption, verified its own breakage as green, burned ~75% weekly quota — 150 itemized failures
- Allow starting a fresh plan file mid-session instead of merging new plans into the existing auto-named file
- [BUG] Cowork/MCP tool calls fail with -32001 after ~60s even when the server completes and the result arrives
- [Feature Request] Add option to hide or consolidate context/token usage indicator in UI
- [BUG] Cowork sandbox bash tool caps timeout_ms at 45000, no way to run longer commands
- Background subagents recursively re-delegate their brief to new background subagents (3-level cascade, duplicate work, misleading completion reports)
- [FEATURE] Unarchiving a Claude Desktop session by editing the session file
- Cowork (Windows): UserPromptSubmit hooks now execute, but stdout/additionalContext is discarded — wire hook output into context at parity with Claude Code
- [BUG] Windows: plugin lspServers bare command (e.g. pwsh, pyright-langserver) refused pre-spawn with 'not found or is in an unsafe location (current directory)' -- breaks official pyright-lsp plugin
- Ctrl+Backspace (delete word) stopped working in prompt input after update to 2.1.199
- Task list sidebar panel for session task visibility
- [BUG] Mid-message user-typed slash command with disable-model-invocation: true is silently ignored despite UI highlighting
- Claude Censorship and the creation of a walled garden around the General Intellect.
- [BUG] Claude code install hangs on a modern Linux kernel on KVM?
- MCP server 'supabase' sendo removido/adicionado repetidamente com comando malformado (openssl rand -hex 32 não interpolado)
- Links open twice on WSL: Claude Code opens via xdg-open (WSLg browser) in addition to the terminal
- Desktop app: assistant text blocks between tool calls intermittently not rendered (answer body missing, only short lines shown)
- [BUG] ← for agents displays even when not in an agents view managed session
- [BUG] SessionStart (and other command-type) hooks fail to execute on Windows — shell resolution drops \bash.exe from Git path
- /model[Feature Request] Anthropic API Safety Filter Error with Model Name "Fable 5"
- [Bug] Intermittent malformed tool-call emission with corrupted "court" prefix delimiter
- [Bug] Git reference resolves to local main instead of origin/main after fetch
- [Bug] Claude 3.5 Fable 5 fails on routine debugging tasks
- MCP server "supabase" sendo removido/adicionado repetidamente com comando malformado (openssl rand -hex 32 não interpolado)
- Live output tail in the /tasks (background shells) panel — inline + pop-out window
- [Bug] Fable safeguards triggered on routine coding work
- [Bug] Incorrect Information About Claude Configuration and Routines
- [FEATURE] Cross-machine session sidebar sync when .claude is on a shared drive
- [BUG] Fable 5 → Opus 4.8 fallback misfires on hobby-OS kernel
- Billing page: unclear flow for adding funds / purchasing credits
- [Feature Request] Make feedback commands more explicit to avoid accidental activation
- [BUG] Desktop app: Code view sidebar stays in English despite ja-JP locale (web is fully translated)
- [BUG] SSH option not available in Claude Code Desktop when using external API provider
- [DOCS] `AskUserQuestion` tool docs still describe v2.1.198 auto-continue as the default; v2.1.200 made it opt-in via `/config`
- [DOCS] New `manual` permission mode value accepted by `--permission-mode` and `defaultMode` is not documented in `permission-modes.md`, `cli-reference.md`, or `settings.md`
- [DOCS] `disabledMcpServers` and `enabledMcpServers` settings docs do not explain the non-array crash fixed in v2.1.200
- [DOCS] Background-agent daemon handover rule now uses the version's embedded build timestamp, but `agent-view.md` and `cli-reference.md` still describe the old version comparison
- [DOCS] `sub-agents.md` API-error section only covers subagents cut off after producing output; v2.1.200 added the "before any text" case
- [DOCS] `claude agents --plugin-dir <dir>` is documented but never warns about flag placement after `agents`
- [DOCS] Project-scoped plugins loading from git worktrees is fixed in v2.1.200 but never documented in the plugin reference or discover-plugins guide
- [DOCS] `/mcp` server list screen-reader focus tracking is fixed in v2.1.200 but never described in MCP docs
- [DOCS] Voice dictation troubleshooting never lists the misleading "Voice connection failed" message fixed in v2.1.200
- [DOCS] Fullscreen `fullscreen.md` tmux caveat still says synchronized output is unsupported; v2.1.200 enabled it for tmux 3.4+
- [DOCS] `axScreenReader` description lists only "decorative borders and animations"; v2.1.200 added glyph hiding, transcript labels, and nested-table-as-`Header: value.` rendering
- [DOCS] Install script's new out-of-memory explanation needs a troubleshooting entry in the installation pages
- [Bug] Claude 3 Opus model produces incorrect outputs requiring self-correction and hallucinations
- [Bug] Fable 5 usage increases without model invocation
- OOM crash under multi-subagent load: subagent CLI processes grow to ~6.5 GB RSS each; CliGovernor logs critical memory pressure for 8 min but evicts nothing (Linux, Desktop 1.17377.0 / CLI 2.1.187)
- "Group by projects" view renders empty on desktop AND claude.ai web - "group by date" works fine
- claude can't self test against self bullshit.
- [Bug] Claude Code CLI truncates long input messages without warning
- Model behavior report (Sonnet 5): overconfident unverified claims, out-of-scope config changes, doubting user's verified failure reports
- [BUG] Contradictory 'session limit resets at X' times across one session; background subagents die terminally and reset projections silently roll forward
- Feature request: group conversations into Projects in the sidebar
- [BUG] Wrong context usage and tokens shown
- [Bug] Excessive token consumption without meaningful progress in extended sessions
- [BUG] Dispatch stuck in infinite self-loop: system repeatedly injects "[Your previous response had no visible output]" after long context
- [FEATURE] Time to complete estimation
- [BUG] Syntax highlighting loses multi-line string state — docstring continuation lines render as code
- Background daemon posts spurious "needs re-authentication" notification when a proactive OAuth refresh fails during macOS sleep/PowerNap dark-wake (transient, self-heals, auth never actually invalid)
- Desktop app: session permanently wedged after manual /compact — deferred local-command send is echo-tracked by a uuid that never echoes
- Desktop: archived sessions are hard to discover — no visible "Archived" filter or section
- [telegram plugin] Orphaned `bun server.ts` poller spins at 100% CPU indefinitely; self-termination guards all fail
- [Bug] Dual-use safeguard false positive downgrades Claude 5 to Opus 4.8 during authorized security review of own codebase
- [BUG] [Model is refusing to help me train a model for CSAM detection[
- [BUG] .claude/settings.json and settings.local.json resolve against literal cwd, not git root — subdirectory launches silently drop all project settings
- Feature request: inline feedback (thumbs up/down) during a session, not just GitHub Issues
- [BUG] Cloud sessions: WebFetch surfaces egress-policy denials as bare 403s — and the allowlist blocks claude.com itself
- [BUG] claude recommends adding non-existent directory to PATH
- [FEATURE] LaTeX rendering support in Claude Code VS Code extension chat/output panel
- macOS: Screen Recording (TCC) permission must be re-approved on every update because the binary installs to a versioned path
- `!` shell commands block on Claude's response before accepting the next `!` command
- [Bug] TODO list persists on screen after task completion
- [BUG] Thinking Spinner Used Inappropriate Word
- Worktree isolation inflates spawned-shell env past ARG_MAX → E2BIG on all Bash, unrecoverable without restart
- Windows: worktrees in git-lfs repos get a stray dev/null/ directory with LFS hooks (hooksPath suppression bypassed via GIT_CONFIG_PARAMETERS)
- [Feature Request] Add customizable ASCII art and animation support for terminal display
- Deeply-nested subagent fan-out causes unbounded memory growth → host-level OOM
- Auto-mode classifier repeatedly flags Agent dispatches as prompt injection due to harness-injected context-mode plugin block
- claude.ai/design: no indication that a project already has an uploaded component library before routing into the from-scratch "Generate" wizard
- Feature request: allow image attachments on the Claude Design feedback form
- [BUG] @claude hello, can you read the README and summarize this project?
- [BUG] @claude please label this issue as "bug"
- [FEATURE] Command line auth browser selection for OAuth login flow (--no-browser / BROWSER env)
- /design-sync skill: reconciliation-delete/no-anchor logic can destroy content added by claude.ai/design's in-app "Generate" step
Resolved issues (315)
- [BUG] Epitaxy (Windows desktop): microphone device selector forgets its choice and reverts to the system default input device -- no way to pin dictation to a non-default mic
- [BUG] Windows Desktop: main process permanently stops writing main.log once it hits the 10 MB rotation cap mid-run
- [Feature Request] Weather Radar Denoiser ML Model Design
- [BUG] Org-level banned flag not cleared after user-level reinstatement — cannot resubscribe to Max (account_banned on PUT /subscription)
- [BUG] Org-level banned flag not cleared after user-level reinstatement — cannot resubscribe to Max, blocks Claude Code CLI workflow (account_banned on PUT /subscription)
- [BUG] Org-level banned flag not cleared after user-level reinstatement — cannot resubscribe to Max, Claude Code CLI stuck on Free tier (account_banned)
- [Bug] Anthropic API: Model-generated text injected into tool_result stream causing silent write failures
- [BUG] Deep Research spawned 105 agents and exhausted Max plan quota in ~10 minutes
- [claude-in-chrome] MCP tab group always named "Claude" — parallel sessions are indistinguishable
- Hardcoded "Open in..." app list in Files panel doesn't respect OS default app
- [FEATURE] Auto Compact
- Allow customizing user message background color in the UI
- remote-control session not found on mobile — redirects to /code/family
- [Bug] gcloud auth login exits prematurely before authentication completes
- [Bug] Tool calls emit malformed format mid-session, causing silent failures with false success claims
- [BUG] [Billing] Pro to Max upgrade fails before issuer authorisation (AU account)
- [Bug] Unable to insert images in agent chat conversations
- Claude Code goes silent when using VS Code Remote SSH Extension and Claude Code on Digital Ocean droplet
- [BUG] Selecting this in WSL has no effect: [Memory Auto-memory: on 1. User memory saved in ~/.claude/CLAUDE.md 2. Project memory checked in at ./CLAUDE.md ❯ 3. Open auto-memory folder ✔]
- [Bug] Prompt injection-like instructions in tool results/session context with destructive commands
- [Bug] Flicker-free rendering breaks Shift+Enter on iTerm2
- [BUG] Gmail + Google Calendar connectors show "Connected" in UI but never trigger OAuth, no Google account linked
- [Bug] Text truncation at right edge in tmux + iTerm2 terminal
- [Bug] Auto-update failed
- [Feature Request] One-shot model override for single prompts without session switching
- [Bug] Fullscreen TUI prevents scrollback in iTerm2
- Project memory & session context orphaned silently when a repo directory is renamed/moved
- [Bug][cyber] Tweaking ffmpeg params to fix too-wide on-screen video aspect ratio wrongly blocked (req_011CcYJyW46wcHoLHF5AJVmu)
- Interactive sessions that inherit CLAUDE_CODE_CHILD_SESSION silently skip local transcript persistence
- [BUG] OAuth login fails on Linux VPS: "Redirect URI platform.claude.com/oauth/callback is not supported by client"
- SSH sessions: history blank and fork fails due to tail connecting via raw IP instead of SSH config alias
- [Bug] Prompt Injection in Sub-agent Output: Attempted Manipulation to Write Fabricated Security Findings
- Agent acts on bundled multi-part approvals; presents decisions as bare issue-IDs; confident stale knowledge
- [Bug] Claude Code auto-update mechanism not triggering
- [BUG] - Sidebar/ruler not visible after latest update , - Unable to scroll up in session, - OS: Windows WSL2
- Make it easier to expand repo access in a Claude code session
- Desktop Code mode creates an orphan empty-context session instead of resuming; UI keeps showing full scrollback
- [FEATURE] Preserve file-change system-reminders across context summarization
- [BUG] claude desktop blows away it's own settings on startup
- [Bug] Anthropic API Error: Server Rate Limiting (Temporary Request Throttling)
- [Bug] Anthropic API Error: Server Rate Limiting During Normal Usage
- /workflows slash command not recognized in VS Code extension
- [BUG] Claude Code shows "out of credits" on overage accounts; requires full system reboot to recover
- Background job writes completion output into the wrong thread — job spawned in Thread A bleeds into Thread B if B is active when the job finishes
- Ultraplan web handoff fails to push PR when container has read-only GitHub access
- [VSCode] Cannot fork a conversation after the latest assistant reply — fork only anchors on user messages
- TUI: click-to-select fires even when the TUI window lacks focus — should require focus-first
- Docs: PreToolUse hook tool_input (Bash) omits run_in_background / description / tool_use_id
- [Bug] Anthropic API: 50% timeout errors causing token waste
- Clickable prompt-summary header → jump to any prior input in scrollback
- Add RTL (right-to-left) layout support for Hebrew/Arabic users
- Tool-call markup occasionally emitted as plain text instead of executed (commands silently not run)
- Paste (right-click / Ctrl+Shift+V) broken in WezTerm on WSL2 — regression from older versions
- Opus 4.8 on Vertex (CLAUDE_CODE_USE_VERTEX) returns empty thinking — display:"summarized" honored on direct but dropped on Vertex
- Schedules tab shows "No scheduled tasks yet" while scheduled tasks run correctly in the background
- [BUG] Android: Question tool drops repeatedly
- [BUG] Claude Desktop (Windows): Bursty frame delivery triggers G-Sync flicker (119Hz → ~40Hz) at end of response streaming and while typing
- [Bug] Inconsistent token usage reporting with similar workloads
- [Bug] Tool fails to apply API after multiple requests
- [BUG] Bash and Read tool outputs arriving one turn late
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [Feature Request] Add redo functionality after undo operation
- [BUG] Multiple issues in new version, working in CLI v2.1.156
- [FEATURE] Allow readline / emacs style shortcuts consistently
- [BUG] API Error: 400 messages.7.content.12: `thinking` or `redacted_thinking` blocks in the latest assistant message cannot be modified. These blocks must remain as they were in the original response.
- [Feature Request] Pin/star sessions in VSCode extension sidebar
- Fork Conversation opens empty session (resume: undefined) in VS Code extension v2.1.157
- [BUG] Usage limit reached at 0% session and 12% weekly; used the in-app support (bot) and they told me to report here as it has persisted across multiple sessions.
- API 400 "thinking blocks cannot be modified" on a fresh session during a multi-tool turn
- [BUG] CC CLAUDE_CODE_DISABLE_1M_CONTEXT doesn't get respected
- [Bug] Claude prints skill contents instead of silently loading instructions
- Approve-once-per-session for repeated identical tool calls
- [FEATURE] Claude in Chrome MCP access Cloudflare Access-protected URLs
- Write/Edit tool fails with special characters in file content on Windows
- Write/Edit tool fails with special characters in file content on Windows
- [BUG] Brave not selectable in /chrome browser picker — reconnect opens & authorizes Brave but it never appears as a selectable browser
- [Bug] Opus 4.8 performance regression: excessive latency with degraded quality and context loss
- [Bug] Skill registration confused with plugin registration in skill-creator workflow
- [FEATURE] Settings menu should default to Usage tab
- [BUG] Up arrow in Chat input skips cursor navigation in multi-line text, goes directly to history
- [BUG] [CLAUDE_CODE_NO_FLICKER] Ctrl-G external editor: raw keycodes logged to terminal, input frozen after editor closes
- [BUG] Desktop "Code" mode: silent "exited with code 15" at startup when a global stdio MCP server is configured (2.1.149+ regression)
- [Bug] Context limit reached during ultracode execution and auto-compact failed
- [BUG] Terminal Mode startup delay: shellIntegration.executeCommand waits ~5s for OSC 633 prompt markers before typing `claude`
- [BUG] Paid Claude Pro but account stuck on Free Plan — Fin AI cannot resolve — Invoice M2R3TACN-0002
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [Bug] Tool results returned out of order with stale batch data when parallel tool calls fail
- [FEATURE] Allow custom skills to appear before built-in commands in the / menu
- [BUG] VS Code terminal output garbled with stray Unicode diacritical characters — resolves after reloading VS Code
- [BUG] Agent fabricated tool output and reported a non-existent prompt-injection "incident"
- Interrupting an interleaved-thinking turn mid-stream wedges the session: unrecoverable 400 "thinking blocks ... cannot be modified"
- Shell snapshot uses unguarded $ZSH_VERSION, breaks set -u (nounset)
- [Bug] Workflow dispatch by scriptPath drops args; user-global ~/.claude/workflows/ not resolvable by name
- Security · `claude mcp add` echoes Authorization Bearer token to stdout — enables credential leak via paste-back verification
- [BUG] Desktop app closes after 1 minute
- [BUG] /usage → Stats → Models, 'Tokens per Day' chart does not render lines in per-series colors
- [BUG] Claude Code fails BASH invocations
- [Bug] Claude Opus 4.8 ignores MCP tool definitions, hallucinates parameters
- Feature request: Organize chats in folders by project + ability to rename chats
- /resume session list is incomplete and not chronologically sorted
- [Bug] Anthropic API Error 400: Thinking blocks cannot be modified on history reconstruction with Opus 4.8
- [BUG] Desktop App: caret can't be placed inside text in a detached session window (Open in → New window) — typed text appends to end of block
- [BUG] /rename replaces the ultracode effort indicator with the session name (effort level stays active)
- [Bug] Safety classifier blocks routine code analysis for paid users (started 2026-05-23)
- Safety Report: AI Guardrails Do Not Work — 56-Day Proof, 06K Loss
- [Bug] Anthropic API Error: Cannot modify thinking blocks in assistant message
- [Feature Request] Make ※ recap: line scannable: break single long sentence, reduce entity density, add structural hierarchy
- [Bug] AskUserQuestion does not block dependent tool calls until user responds
- [MODEL] opus 4.8 Claude Code autonomously consumed user API credentials without approval
- [BUG] indows 11 Pro: Cowork, Connectors, and Claude Code all broken after latest auto-update — ERR_CONNECTION_RESET + OAuthError
- [BUG] [Opus 4.7] Korean lexical fixation: repeated insertion of "영역" in unrelated output
- query() with image content block via AsyncIterable[dict] prompt → 400 "Could not process image"
- [Bug] Explore agents hitting context limit with read-only tools
- [Billing] claude.ai Max (20x): ~$400 in usage-credit auto-reload charges against <10% plan usage, no overage notifications
- [BUG] Long-session compaction loses operator-action attribution → model frames operator-driven Bash runs as passive/autonomous
- [BUG] Reducing bugs in agentic coding (from a real money-path build)
- Doing useless tool call wasting token
- [Bug] Opus 4.8 Fails to use tools and softbricks context: Cannot modify thinking blocks in assistant message
- [BUG] Sonnet repeated violated explicit guidelines about commands not to execute
- Tool-call XML emission silently drops parameters on bare tags / unescaped closing-tag substrings (diagnosis misrouted to receiving end)
- [Bug] Anthropic API Error: Cannot modify thinking blocks in assistant message
- [Bug] Tool-call parser rejects indented XML tags with vague "malformed" error
- [Bug] Anthropic API Error: thinking blocks in assistant message cannot be modified
- # [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
- Claude omits modified files when staging commits with new files
- Allow running shell commands during active turn without canceling
- [BUG] Up/Down arrows in input box navigate history instead of moving cursor — regression in 2.1.149+
- Feature Request: Cross-device session sync between desktop and mobile
- 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
- [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)
- Better usage model (pay to push to reset)
- [Bug] Anthropic API Error: Server Rate Limiting on Recurring Requests
- [Bug] Claude hallucinating code without repository context validation
- [BUG] /usage command shows unusal UI
- AskUserQuestion overlay should be minimizable / dockable — it blocks the message it refers to
- Hook stdout shown + suppressOutput:true ignored in 2.1.158 (regressed from 2.1.157)
- [Bug] advisor() silently returns no result (red circle, ~4s, no advisor_tool_result) when transcript exceeds the 100-image API limit
- Auto-mode classifier: `allow` rules silently ignored for default-branch protection; self-modification block contradicts documented escape hatch
- [BUG] terminal-notifier 桌面通知在权限弹窗时不触发
- [Bug] Parallel tool call execution cancelled on Bash error
- Session permanently wedged on 400 "thinking blocks cannot be modified" after parallel tool_results
- [BUG] Remote Control (Android version) always fails to deliver .APK files
- [Bug] Tool output replay corruption with batched parallel calls causes phantom state
- [BUG] mcp__claude-in-chrome__* tools are advertised but not callable after upgrade to Claude 4.8
- [BUG] Built-in local slash command output (e.g. /context) doesn't create a rewind checkpoint, while plugin commands do
- Feature request: persist collapsed/minimized state of Preview log panel
- Workflow: transient "Connection closed mid-response" in one agent aborts the whole run and discards all completed work
- Artifact tool: allow setting sharing/visibility programmatically
- I appreciate you reaching out, but this message is about a WiFi connection issue unrelated to Claude Code, the Anthropic CLI tool. If you're experiencing issues with Claude Code specifically, please provide: - What command you ran - The exact error messag
- AskUserQuestion choices are clickable with mouse — accidental clicks hard to avoid
- Ctrl+Backspace (delete word) stopped working in prompt input after update to 2.1.199
- Claude Censorship and the creation of a walled garden around the General Intellect.
- [BUG] cannot access appeals form against nondescript claim
- [BUG] Greyed out Figma installed button - can't push design from Claude to Figma
- Google Drive MCP integration missing write/edit support for existing Docs
- [Excel Add-in] Keystrokes go to active cell instead of Claude input box; cell chip reappears after dismissal
- [Bug] Claude Code generates incorrect content instead of performing language corrections
- [Bug] High task failure rate (82 failures) with basic operations
- [Bug] False positive detection occurring repeatedly
- [Bug] High error rate in task execution: 59 violations and 17 wrong deliveries 1 TASK
- [FEATURE] Let the agent programmatically create and manage conversation threads
- [BUG] Remote Control incorrectly queues msgs with no way to send/clear from the remote side (only local Esc recovers)
- Bug: /remote-control fails silently on an outdated / non-auto-updating build
- [Bug] Claude-in-Chrome screenshot capture hangs on document_idle timeout despite page being idle
- [Bug] Ctrl-O exits batch processing and cancels all pending operations
- [Bug] Auto-update failed
- [BUG] Custom statusLine permanently stops being invoked after a subagent/response completes at project scope — but not at user scope
- MCP channel push: non-blocking delivery mode for active sessions (multi-agent coordinator use case)
- ultrareview consumed a credit despite crashing before producing any findings
- [Bug] Tool Call Parsing Error: Unparseable Model Response (Retry Failed)
- /remote-control regressed: now blocked in Claude Desktop's local-agent (non-interactive) session
- App Build with Terrible Results
- [BUG] API 400 error: thinking/redacted_thinking blocks cannot be modified after Claude Opus 4.8
- Auto-mode permission classifier over-blocks an explicitly-authorized tool call in a multi-task prompt
- Tool outputs expanded by default — no setting to collapse, inconsistent across machines
- [Bug] Channels plugin (Telegram) not functioning on Linux - notifications not injected into session
- [BUG] GitHub MCP server repeatedly disconnects during web sessions, wasting tokens with no output
- Stop hook does not fire when the CLI is terminated by SIGTERM (e.g. external execution timeout)
- [BUG] Gender Bias
- SIGILL crash on Intel Westmere (Xeon X5675) — native binary requires AVX2SIGILL crash on Intel Westmere (Xeon X5675) — native binary requires AVX2
- [FEATURE] add startup/session-init hook event to settings.json hooks
- [BUG] My claude keeps exhausting up my claude limits
- [BUG] Stopped due to usage limits being reached, but I’m only at 64%.
- Claude Code recommended services without verifying API availability, causing wasted spend
- Claude fabricates data presentation and produces inconsistent outputs in same session
- [BUG] Claude in Chrome `file_upload` rejects all scheduled-task sessions with misleading error (real cause: INVALID_SESSION)
- Multi-plugin custom marketplace: only first plugin registered in installed_plugins.json, skills don't load
- [BUG] tree-kill dependency causes pgrep storm / 100% CPU when reaping Bash subprocess trees on macOS
- [BUG] Scheduled routine runs ignore configured model and fall back to Sonnet 4.6 — manual runs use correct model (Opus 4.8)
- [Bug] Anthropic API Error: Invalid tool input_schema type during compaction
- [Bug] Session limit blocks access despite available credits and low usage percentage
- Malicious shell command injected into assistant-generated tool_use (command field corrupted with forged result-wrapper tags)
- [Feature Request] Make worktree/branch label truncation in info bar configurable
- [BUG] `claude mcp get` prints full env values in plaintext (secret leak)
- [BUG] Left arrow in AskUserQuestion doesn't return to Agents view
- [Bug] Parallel tool batch cancels all remaining calls when one call fails
- [BUG] Safari browser fails at login into Claude account via Google SSO
- Desktop app: diff viewer shows stale content after Claude edits files
- [FEATURE] Surface subagent-generated plans in the Desktop Plan panel
- [BUG] VS Code extension blank panel — this.webviews.add(...) is not a function in setupPanel (VS Code 1.122+)
- [BUG]
- [Bug] Stop hooks fail with "node: command not found" and missing plugin directories
- [BUG] terminalProgressBarEnabled (OSC 9;4) doesn't work inside tmux
- [FEATURE] Allow `mcp_tool` SessionStart hooks to inject their result into context
- Multi-lane Workflow agents re-create shared prompt cache per lane (no cross-sibling sharing), causing excessive token usage
- [BUG]
- [BUG] Repeated Bash tool-call loop during git workflow consumed session limit
- I don't see a bug report for Claude Code in your message. This appears to be a conversation about macOS screen recording file behavior and a discussion about communication patterns. If you're reporting a bug with Claude Code, please provide: 1. What you
- Feature request: paste image from clipboard directly in terminal (Ctrl+V)
- [BUG] dispatch session fantome
- Plugin lifecycle gaps observed while building a statusLine plugin
- Max 20x: 5-hour rate limits not doubled despite May 6 announcement
- [BUG] /fast has no effect in desktop app — ON and OFF both ~11s; terminal ON ~2x faster (~6s)
- [Bug] Anthropic API Error: Usage credits required for 1M context window
- Claude Code shell does not inherit AWS_REGION from user's terminal session
- [Bug] Agent avoids Edit tools causing duplicate calls and rapid context exhaustion
- [BUG]
- [BUG] Unable to copy `%PDF-1.6` text
- [FEATURE] Multiple account
- [BUG]
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- macOS: keybinding hint shows 'meta+w' instead of 'Option+w' (⌥W)
- [BUG] Persistent server-side session billing phantom usage — cannot be killed from client
- Auto mode selected at plan approval persists after execution completes
- Enriched input signal for Claude Code - Eyeballs for non-verbal communication cues
- [Feature] Add a user-facing toggle for spinner verbs / tool-use summaries in Claude Desktop
- [BUG] agent stuck: git-rebase preference + auto-mode -> blocked force pushes -> requires human intervention
- Support spinnerVerbs customization in desktop app
- Agent tool model parameter should accept full model IDs, not just tier names
- [BUG] CLAUDE_CODE_DISABLE_1M_CONTEXT ignored in 2.1.156
- [FEATURE]
- [Bug] Duplicate output in terminal causing content duplication
- [Bug] Auto-accept mode ignores directory containment check for home paths ending in period
- Windows: background daemon keeps a directory handle after `claude rm`, blocking directory deletion
- [BUG] LSP client never sends textDocument/didClose, leaking language-server resources until OOM
- Ordered-list markdown in submitted messages gets auto-renumbered (e.g. '2.'/'5.' → '1.'/'2.'), clobbering intended numbering
- [FEATURE] Per-banner "Don't show again" dismissal for MCP connection notifications
- Permission/question dialogs do not render in Ctrl+O transcript mode (UI hangs indefinitely)
- [BUG] claude recommends adding non-existent directory to PATH
- [Bug] Excessive token consumption per prompt with Opus 4.7 1M context model
- [BUG] Opus 4.7 excess token usage in Windows - 7K tokens used 52%
- [Bug] Tool-layer unreliability in long/compacted sessions: duplicated results, delayed delivery, empty/garbled file reads
- `acceptEdits` mode does not extend to `additionalDirectories` (Edit/Write still prompt per-file)
- [BUG] API Error: 400 — thinking or redacted_thinking blocks in the latest assistant message cannot be modified.
- [BUG] MCP tool parameters with type: number are sent as strings to MCP servers
- [BUG] Cowork: uploaded context lands in the project mount, not mnt/uploads/
- Parallel tool-call batch cascades "Cancelled" to independent calls when one fails
- Opus 4.8 / xhigh / Claude Code: model asserts specific tool-output values BEFORE the tool calls return — model self-diagnoses the bug in-context but keeps doing it
- Benign `node -e` regex script false-flagged as "Fork bomb detected"
- Edit tool silently substitutes Unicode smart quotes for ASCII double quotes in shell scripts
- Claude Code context window not resetting — stuck at 1M instead of 200k on Pro
- [BUG] 2.1.153 silently discards tools/list response from rmcp 0.12.0 HTTP MCP server (works in 2.1.152, wire-identical handshake)
- [BUG] Cowork shows "yukonSilver not supported (status=unsupported)" on Windows 11 Pro with 11th Gen Intel i5-1155G7 — VM never initializes
- Tool results delayed and duplicated in interactive sessions (v2.1.158, Linux, multiple concurrent sessions)
- Intermittent tool-result corruption: stale/replayed output + sibling cancellation (v2.1.143)
- [BUG] MCP: structuredContent overrides content field, losing actual tool response
- [Feature Request] PostToolUse hook で output 文字列を改変 (REDACT) する機能の追加
- [Bug] Opus 4.8 hallucinating security incidents and fabricating evidence during long tasks
- Workflow counts API-errored subagents as "completed"; 429/500/529 storm under multi-agent load
- [FEATURE] Google Vertex support for Remote Control features
- Server-side tool (server_tool_use) emitted but no result returned; response reports stop_reason: end_turn
- Token waste bug: Workflow spawned 44 parallel agents to read files instead of using git clone (2M tokens vs ~1000)
- [BUG] Installer places binary in VS Code Snap directory, leading to silent uninstallation on Ubuntu
- [BUG] Windows: PowerShell tool absent from schema when Git Bash is on PATH, despite pwsh.exe being present and detectable
- [MODEL] Porting JS to TS Relies Too Much On Feedback From TSC
- [BUG] Remote Control: "is not yet enabled for your account" persists for Max user across all documented troubleshooting steps
- MCP: re-enumerate tools mid-session for remote-control sessions (mcp_reconnect works but is unreachable)
- Agent deflects with "untouched by my change" after its own change breaks previously-passing tests — needs accountability-first framing
- Bug: Remote Control via remoteControlAtStartup has no persistent indicator — startup notice scrolls off, nothing in status line / corner
- [BUG] Single README.md edit burns entire 5-hour Max 5x window in 9m 39s (Opus 4.7 1M, resumed session, v2.1.126, JetBrains, Windows)
- [BUG] Thousands of ghost "root-server" sessions regenerate on every launch, crashing the app — survive full uninstall/reinstall and all known data clearing
- In-flight /btw side conversation silently lost on Code -> Cowork -> Code switch (data loss)
- Display `~/` instead of full `/Users/<name>/` in tool call path labels
- [Bug] Model version inconsistency across devices with same account, requires reinstall
- [FEATURE] Dynamic Workflow Feature List
- [Bug] VSCode terminal output displays corrupted text with garbled symbols
- [BUG] Tool-call results empty in live UI then flush late/out-of-order (Bash, Read, MCP, advisor) — macOS, Opus 4.8, parallel batches
- [BUG] Review crashed before producing findings
- CLI surfaces transient 429 "rate_limit_error" to user with no backoff/retry; session sits idle until user types `continue`
- [BUG] Abnormal Session Limit Drain Since Opus 4.8 Rollout (May 28, 2026) — VSCode Extension, Pro Plan, Windows
- opus-4-8 appears to re-fetch identical tool results after normal (non-error) returns — 2-3x vs opus-4-7
- [MODEL] Claude Code + Opus 4.8 frequently echos hello world during planning
- Model autonomously deployed to production without user authorization
- [BUG] Windows + Git Bash: tengu_cobalt_ridge experiment silently force-enables the PowerShell tool & "Shell: PowerShell", causing wrong-shell retries (token waste)
- SessionStart hooks crash when dispatcher routes through prompt-hook path expecting ToolUseContext
- Opus 4.8 burns 5h limit in ~30min on operational work; no signal to downgrade to Sonnet
- Parallel tool calls: one benign tool error cancels all sibling calls → agent misreads as broken shell and flails
- Conversational-deferral (patcher) behavior evades anti-defer hooks: file-content scope + ≥2-occurrence threshold blind spots
- [BUG] Routine scheduler skips Monday when timezone is UTC+8 (SGT/CST) — weekday check applied in UTC instead of local time
- [MODEL] Korean output corruption — words replaced with "영역" token in long responses
- [FEATURE] Multiple subscription account
- [FEATURE] Jetbrains plugin should support setting parent folder as working folder
- [Bug] Frequent tool execution errors with Claude 4.8 Opus model
- autoCompactEnabled does not trigger automatically despite being set to true
- Permission dialog: make button order configurable
- [Feature Request] Sidebar: nest sub-folders within the same Git repo as separate project blocks
- Opus 4.8 Model stated multiple unverified claims as fact and fabricates specifics within a single session
- [Bug] Claude 4.8 model stuck in infinite loop executing spurious bash commands
- [BUG] iPad: app crashes (SIGABRT) when tapping "Default" environment picker — UICollectionView invalid scroll index path
- illegal hardware instruction" on Intel Core i5-3470S (Ivy Bridge, no AVX2) - crashes on startup
- [Bug] Workflow agent marked complete on Anthropic API policy violation error
- [Bug] Anthropic API Error: Cannot modify thinking blocks in assistant message
- Ability to delete conversation portions from context messages
- [Bug] Agent dispatch consumes excessive tokens without course-correction mechanism
- Bash/Read tool results intermittently return empty (then flush late, redundantly) on Linux — NOT length/concurrency-correlated; same symptom as stale-closed #36038
- [BUG] Claude in Chrome v1.0.74 — MCP bridge navigation and read tools wedge after May 27 update
- [BUG] Sub-agent dispatched via Agent tool stalls silently on MCP permission gate — no surface to parent CLI UI
- Chinese IME input: characters overwritten/dropped when editing mid-text
- [BUG] Linear MCP OAuth /authorize endpoint redirects to deprecated gdrive install URL
- Conversation permanently blocked after credential-related context — rewind doesn't help
- Tool-result transport desync/replay/fabrication in claude remote-control daemon under large parallel tool batches
- [BUG] --resume normalizes model ID, breaking custom API endpoints that require the original model name