Digest / October 6, 2026
Claude Code Issues Digest — October 6, 2026
276 new issues filed ·
156 resolved in anthropics/claude-code (UTC day).
New issues (276)
- [BUG] Desktop Code tab: plugin Select/Input labels are read twice by VoiceOver
- [FEATURE] Desktop app (Code tab): jump from a quoted reply back to the quoted passage
- [BUG] Cloud scheduled tasks em modo auto: o classificador de permissao bloqueia acoes autorizadas pelo dono, e nao existe turno humano para aprovar
- Agent failure: treated a report's timestamp as the event's time
- Agent failure: buried the user in process noise instead of moving the goal
- [GitHub integration]
- Agent failure: fixed a different problem than the one asked and reported it done
- [GitHub integration]
- [GitHub integration]
- [FEATURE] Prefill the rename input with the current session name
- [BUG] Windows: VS Code extension ignores project permissions.allow because workspace trust is keyed by drive-letter case (c: vs C:)
- [Feature Request] Add cybersecurity lab bypass allowlist for educational contexts
- [BUG] "Waiting for API response · check your network" during thinking is a 60 s stall in the thinking-summary step (showThinkingSummaries: true), no network problem
- [FEATURE] Desktop/Claude Code: stdio MCP servers duplicated per session (~1 GB each) and can't be started mid-session
- [BUG] CLAUDE_CODE_EXTRA_BODY thinking field is merged into WebSearch/WebFetch requests and breaks them (still on 2.1.290; follow-up to #56984)
- [BUG] v2.1.290: --resume on opus-5-5/sonnet-5-5 re-writes the whole history to the prompt cache every time (haiku unaffected)
- Desktop app: 'auto mode classifier' refuses supervised Claude-in-Chrome actions while the session is in bypassPermissions
- [Bug] Voice mode truncates input text on large messages without recovery
- [FEATURE] Claude Code Group by folder is misleading because it groups by repo. Add option to actually group by folder.
- [Bug] Anthropic API Error: 403 Access to this model requires an access grant
- [Bug] macOS Gatekeeper rejection + TCC permissions reset on every Claude Code update
- [security-guidance] Commit-review findings replaced by "claude.ai connectors are disabled" warning in the rewake notice
- [security-guidance] Reviews that fail with API errors (e.g. HTTP 401) are reported as clean and marked reviewed
- Claude Code blocks explicitly authorized hardware monitoring setup on my PC without explaining why
- [GitHub integration]
- [Bug] Unclear error message when user is blocked or denied access
- [FEATURE] iPhone and web app interactivity to match terminal suggestions
- [BUG] Dev servers started by Claude are hard to find and outlive the app
- Bash tool grep shim: -q/-c/-l and output to /dev/null report no match when stdin has one invalid UTF-8 byte, while printing lines works (macOS, 2.1.290)
- [BUG] Plugin Button loses a quick second click: it selects the label's text instead of pressing
- [BUG] Cowork: prompt change on a "Require this computer" scheduled task fails under "Skip all approvals" (needs_device_approval, no card); "Automatically approve" works
- OAuth refresh failure caused by a transient network error logs the CLI out permanently (daemon / Remote Control on an always-on Mac)
- [GitHub integration]
- [Windows] Each prompt spawns 3–5 git.exe processes that each open a visible console window
- [BUG] Max Subscriber Account Fully Closed / Broken Appeal Loop after activating $250 Claude Code Cloud Session Credit
- [BUG] Linux: PID namespace/procfs mismatch breaks PDF/.ipynb Read and single-file Grep, and can return another file's content
- Thread hung and became unresponsive.
- [BUG] Worktree guard refuses git -C ~/… in user ! commands; reason withheld under Remote Control
- [GitHub integration]
- [BUG] security-guidance: model reviews 401 behind an LLM gateway because ANTHROPIC_CUSTOM_HEADERS is never sent
- [BUG] Desktop Projects: thread loses auto-approve after the device bridge reconnects; every command then needs manual approval and there is no mode control to restore it
- [BUG] Bedrock "400 Input is too long." (oversized image payload) isn't treated as request_too_large, so the session is stuck for good
- [BUG] Auto-compaction runs without user consent and degrades quality of long sessions
- [GitHub integration]
- [BUG] Desktop app: session opened from a spawn_task card starts at medium effort, ignoring effortLevel in settings.json and the parent session's effort
- [BUG] Mods: prompt box draws the built-in edit before a prompt.edit hook's answer
- always having problems linking my computer
- [Desktop] WebFetch still prompts "Allow once" in auto mode despite a bare `WebFetch` allow rule
- Exchange counts a wallet's fresh fill twice for 1–3 seconds and rejects its next order
- AI do not stop. when i made the typo so i tried to stop the ai. the problem was…
- [BUG] Desktop split view: plugin AbovePrompt band renders only in the leftmost pane
- [GitHub integration]
- the details are showing up in other chat window but not in the project
- [Bug] Opus generates repetitive agent loops and produces faulty tests requiring multiple edits
- [Bug] Login Authentication Failure on Local Machine
- [Bug] Anthropic API Error: Safety classifier blocking file creation for approved commands
- Style the "Transcript saving is off — CLAUDE_CODE_SKIP_PROMPT_HISTORY is set" line as a neutral status (like "bypass permissions on"), not a warning
- [Feature Request] Add history navigation for /btw command responses
- [BUG] prompt.suggest hooks never run for the engine's own suggestion in SDK mode (Claude Desktop Code tab)
- [Bug] Claude extension repeating output in loop after update
- [Bug] Session blocked for legitimate CCTV stream processing and ML analysis use case
- [Feature Request] Support multiple IntelliJ project workspaces and context roots
- [BUG] **Title:** VS Code extension 2.1.291 fails to register commands on WSL - "command 'claude-vscode.sidebar.open' not found"
- [BUG] Scheduled task ("routine") model setting silently reverts from Haiku to Sonnet between unattended cron runs
- [Feature Request] Add vulnerability scanning capability with Fable model support for CVP-verified organizations
- [Feature Request] Add delete session button to /resume tab
- [FEATURE] Allow headless (claude -p) sessions to publish and update Artifacts
- [FEATURE] Show a one-line dictionary definition next to the spinner verb
- [Bug] Auto-classifier executes block action without providing reason
- [FEATURE] Parent session gets no signal when a background subagent is blocked on a permission prompt — unattended runs stall for hours
- [GitHub integration] Repo linked via "Add from GitHub" in chat, but Claude gets no GitHub tools and cannot read it
- The coordinator cannot access my pc locally. He gives me prompts to copy like i…
- [GitHub integration]
- [BUG] claude.ai/code redirects to claude.ai/code/onboarding
- [Bug] Voice input fails silently after auto-update due to broken process attribution (macOS)
- [BUG] Spelling error in claude code UI
- False-positive deletion check: read-only compact Bash command flagged as running rm (2.1.291)
- Worktree-isolation guard refuses plain git commands after a PreToolUse hook wraps them in a launcher (rtk)
- Claude Code inconsistently includes my co-author attribution
- [Bug] Claude Code doesn't request clarification for missing context in user requests
- [FEATURE] create_scheduled_task / update_scheduled_task cannot set the routine's folder (cwd) or worktree isolation
- Dangerous-rm prompt raised by a sub-agent has no auto-deny timeout, stalling unattended bypass-mode runs indefinitely
- [GitHub integration]
- [FEATURE] Claude Desktop - Continued Voice Dictation
- VS Code extension: spinner glyph ignores reduced-motion settings
- [FEATURE] Inline code completions (ghost text) in the VS Code extension
- Withdrawn
- [FEATURE] Function hooks: include the status line's prompt_cache figures in $.session.usage() and session.measure
- [FEATURE] Per-model breakdown of the session cost in $.session.usage() / session.measure and the status line JSON
- [BUG] Organization disabled - appeal TS-01a10b45-3520-73ce-abca-bf46f52162e4 ended in "final decision", requesting re-review
- Agent refuses to type test-account credentials into login forms of a company dev/test environment, inconsistently across sessions
- I cannot see the thread plan on a different computer. The thread plans do not s…
- [FEATURE] Plan mode: cycle between full plan, diff only, and highlighted view with arrow keys
- Auto mode classifier denies a local file edit on the agent's own branch as [Modify Shared Resources], even after explicit user instruction
- [GitHub integration]
- [GitHub integration]
- [Bug] Tool calls fail repeatedly requiring rewind to recover
- [Bug] Claude Opus 5.5 inconsistently refuses security investigation after mentioning suspected bugs
- [Bug] @ mention picker not functioning correctly
- quran[GitHub integration]
- [FEATURE] Agent View as an alternative to Project View
- [BUG] Claude Code usage is consumed while Claude Code is completely unused credits!
- Expose a continued/forked session's predecessor id to hooks and mods (and document the continued-in record)
- [BUG] Claudia ignores CLAUDE_CONFIG_DIR and reads credentials from ~/.claude
- [GitHub integration] Github shows it is connected to Claude but when asked to check a repo ist says it cant and its private
- [GitHub integration]
- [BUG] Desktop app (macOS): Japanese IME stops composing while typing fast; keys are inserted as Latin letters though the input source stays Japanese
- [Bug] Untracked file "forge/Immagine.png" preventing operations
- ik zie een slotje en kan daardoor niet threads verplaatsen naar dit project. Ho…
- [Feature Request] Clarify Claude Code purpose and capabilities in documentation
- [Feature Request] Provide configuration options for security filter sensitivity levels
- [GitHub integration]
- [Bug Report] Invalid Input: Unable to Process Non-Bug Report
- Background daemon: bg-spare workers survive upgrade restarts, low-memory gate ignores cgroup limits, claimed workers indistinguishable from spares
- Subagent budget of 30,000 tokens (new in 2.1.291) is not configurable and not documented
- [BUG] Desktop app (Code tab, Windows) – Markdown file viewer ignores link position in preview mode
- [BUG] `claude mcp serve`: Bash run_in_background output is unreachable since TaskOutput was removed (2.1.277)
- [BUG] Markdown table card fallback wraps wider than the transcript when a mod pane is docked
- [FEATURE] Let a hook read the session's task list, through its input or a documented file format
- [FEATURE] Support negotiated Org/Portable Org output formats
- [BUG] Session transcripts and prompt history vanish while other session files survive
- [BUG] VS Code: "Resume session" on a conversation that entered a worktree starts a fresh empty conversation instead
- [Feature Request] Add concise technical documentation mode for Claude Code CLI
- [BUG] Max 20x: Fable 5.1 refused and billed to usage credits once the general weekly limit is hit, while the separate Fable weekly bucket shows ~90% unused
- [Feature Request] Inconsistent session grouping between remote-control CLI and Claude Desktop
- Quota-aware effort pacing
- [GitHub integration]
- [GitHub integration]
- [GitHub integration]
- [Feature Request] Restore full UI testing capabilities with expanded tool permissions
- [Bug] Model downgrade to 4.8 when using claude-opus-5.5 with date range prompts
- [BUG] Server-gated 100k-token sub-agent budget (2.1.290+) makes delegated work impossible; enabled for one account × Fable 5.1 only
- [Bug] Server-gated token budget feature causes sub-agents to prematurely exhaust context and create infinite respawn loops
- [Bug] Agent obscures or hides incorrect responses mid-execution
- [MODEL] Opus 5.5: overcomplicated, self-contradictory answers and unverified guesses
- [FEATURE] Show where each session is stored (cloud vs. local-only) in the desktop app sidebar
- [BUG] PR badge and statusline `pr` field never appear for GHE.com SSH remotes with a non-"git" SSH user
- [BUG] Drop HTTP connection on "access denied" API errors
- [FEATURE] Plugin UI API: add text direction (dir / textAlign) to Box, Text and Markdown so mods can render RTL (Hebrew/Arabic)
- [Bug] Tool decisions not being persisted or reverted unexpectedly
- [GitHub integration]
- Plugins synced from claude.ai can't satisfy a declared plugin dependency, so the dependent plugin is disabled
- [Bug] Incomplete Bug Report: Missing Details
- [FEATURE] [Desktop] /soft-clear: clear context but keep chat history (in session)
- [Bug] Completed Anthropic Interviewer survey still shows invite prompt on startup
- [Bug] Unclear error flagging on present-tense capability questions
- [FEATURE] Let another local process see and answer a running session's pending permission prompt
- [Feature] Let an SDK / -p host declare its session attended, so background commands aren't stopped at 30 minutes (as in VS Code since 2.1.288)
- [GitHub integration]
- [Bug] Mouse click on plugin Button never fires onPress in fullscreen TUI
- [BUG] # Cowork (Windows): local scheduled task runs no longer get the local artifact tools (list_artifacts / update_artifact / verify_artifact) since 2026-09-30
- Preview pane stuck on blank New tab until clicking Open in chat card; reload doesn't fix it
- [Bug] Unexplained stubborn behavior in command execution
- pretty much every single archive that this thread delivers is incorrect and it …
- delete artifact approvals are not opping up to allow me to approve
- [GitHub integration]
- i see sometime there is a communication drop between the coordinator and thread…
- [FEATURE] Mods: a hover delay for Box hover reveals in Desktop
- [Bug] Sandboxed shell commands freeze typing when executed in directories with many files
- [BUG]
- [Feature Request] Improve visual consistency and animation smoothness for chart rendering between frames
- [MODEL] Orchestrator overrides the user's explicit decisions and repeats rework he said to stop
- nao está a responder o projeto... por mais que tente..
- I fucking hate this update, i want to go back how it was. get me off of this sh…
- Withdrawn
- No se ha podido conectar con tu dispositivo
- [FEATURE] URGENT Add Autoreviewer like Codex for Permissions Requests
- Text "GOK" appears in the prompt at startup in iTerm2 (kitty graphics query reply not consumed)
- [Bug] Model performance regression: Unexpected degradation in output quality
- [GitHub integration]
- [Bug] Misleading error message hides permission error and suggests harmful system-wide file descriptor limit
- Session name loses its zero-width joiner in the session registry
- [FEATURE] Show tool inputs in full while keeping tool outputs truncated (separate display control for input and output)
- [Feature Request] Per-model usage credits restriction to prevent overage on plan-included models
- [GitHub integration]
- [BUG] Desktop: empty custom sidebar groups vanish from the sidebar after restart but stay stored, so recreating them piles up duplicates
- [Bug] Security alerts triggered during development with protected application
- claude plugin test: test $ lacks nouns the plugin adds in engine.create
- Let the agent clear its own context after saving its state to a journal, then continue from that journal
- [GitHub integration]
- [GitHub integration]
- `claude plugin details` reports `Agents (0)` when plugin.json lists agents by explicit file path
- [FEATURE] Back up session transcripts and let resume and fork recover from the backup
- J'ai demandé une simple relecture de script qui m'a couté 40% de mon quota hebd…
- In a private project, Claude told me my Mac was offline when it wasn't. On Oct …
- [BUG] Messages queued during a running turn are lost when the session exits and is resumed
- [BUG] Resuming a session interrupted during a usage-limit retry wait answers the pending message with a synthetic "No response requested."
- [BUG] --dangerously-skip-permissions (and bypassPermissions mode) does not suppress Write/Edit approval prompt in --print mode
- Response Information Keeps Deleting
- In a Claude Code Project, the Claude desktop app identifies my connected Mac as…
- PreToolUse hooks declared in project .claude/settings.json never fire (user-level works)
- [BUG] Dictation (record button) does nothing when the default audio output device cannot start
- [Bug] Goal Stop hook loops infinitely when final step requires user-only authentication
- Put the session's model in every hook's input
- Claude in Chrome: „Allow all actions on <site> for this session“ hält nicht, jede Aktion fragt erneut
- [BUG] Resumed teammates stay bound to the previous team after lead restart; SendMessage to team-lead reports success but messages are never delivered
- New claude projects format is generally good but massively hard to use with two…
- false-positive na "cybersecurity" promt to siema masz...
- [BUG] Claude gateway rejects builtinBrowserEnabled in policy desktop block
- [Bug] Fable advisor usage not counted toward Team plan weekly Fable limit
- Model memory and conversation state as a shared Git-style repository, with all surfaces as clients
- Agent checks in and asks questions instead of completing the assigned task (user feedback treated as discussion, criticised version shipped)
- [Bug] Legitimate requests blocked by overly restrictive content filtering
- [Cowork web] Safeguards repeatedly halt routine tool calls mid-reply in one session (upload, issue filing, session rename)
- I appreciate the feedback, but I'm not sure this is a bug report for Claude Code. Your message seems to be general praise rather than a technical issue to report. If you have an actual bug or feature request for Claude Code that you'd like me to create a
- [BUG] VS Code extension Shift + Drag and Drop file in chat window not working
- [Bug] Safety classifier inconsistently blocks tool calls from subagents vs main session
- [Bug] Claude Code truncates or flags research-related queries
- [Bug] Claude Code not following user instructions and withholding information
- [Desktop/Windows] SSH remote sessions unusable on TOTP 2FA hosts: each connect opens 3-6 separate SSH logins needing fresh one-time codes
- [GitHub integration]
- [Bug] Undocumented breaking changes to exercise metrics calculation and UI display logic without approval
- [BUG] Plugin Pane: an absolute Box with overflow="hidden" draws its clipped content near the window's top after scrolling past it
- [BUG] Auto mode classifier denies an Edit to a Markdown file as [Irreversible Local Destruction] when the prose being edited describes file operations
- [Desktop] Relative file links break after Bash `cd`: model uses new cwd, file pane uses session folder
- Workflow: triggering user message is relayed to every node agent with authority above its computed task — mechanical nodes answer the question instead
- Workflow resume replays cached node results that are transient or that crashed the script
- [MODEL] Orchestrator's fixes left version/status wording on all 35 pages of a site until after the deadline, and its own brief added more
- [MODEL] Orchestrator asked a question instead of building user feedback, shipped the version the user had just criticised, then attempted an unrequested delete
- [MODEL] Workers reported 'live parity' and 'done' at 15:43 while every page still carried the user's forbidden wording
- [MODEL] Orchestrator launched a six-episode audio re-record after the user had decided to stop regenerating audio; 51 published versions in two days
- [MODEL] Agent briefs/replies refused by the user's own hook: NO RELAY (15 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: NO REDIRECT (7 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: SITE MISSING (6 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: JOB MISSING (5 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: JOB DUPLICATE (4 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: NO EDIT (3 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: THROTTLE (3 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: DECISION LOCK (3 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: READ-FIRST MISSING (2 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: NO STOP (2 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: FAKE REPLY (1 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: GUARD UNTESTED (1 on one day)
- PreToolUse hooks read a transcript that lacks the current assistant message until the tool result lands
- clode [GitHub integration]
- [MODEL] Agent briefs/replies refused by the user's own hook: SITE DUPLICATE (1 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: READ-FIRST BLOCK (tool use before reading the page) (26 on one day)
- [DOCS] Update stale docs.claude.com documentation URLs in issue template config.yml
- [MODEL] Agent briefs/replies refused by the user's own hook: DESTRUCTIVE ACTION BLOCKED (false positive) (4 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: UPLOAD BLOCKED (worked as designed) (1 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: STOP BLOCK: reply uses internal ids (8 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: STOP BLOCK: reply hands the user a task (5 on one day)
- [MODEL] Agent briefs/replies refused by the user's own hook: STOP BLOCK: release gate (worked as designed) (2 on one day)
- 1
- Sandbox/permission behaviour changes in patch releases aren't in the CHANGELOG — headless claude -p integrations must re-validate every update
- [BUG] Windows: the Bash tool halves every backslash pair in a command before Git Bash parses it (heredocs and single quotes included)
- [MODEL] Summary: one day of agent behaviour in a multi-agent Claude Code session (23 linked reports, 99 refusals by the user's hooks, one hook that failed open)
- `claude -p` misses the prompt cache after a turn without thinking (Opus 5.5, Fable 5.1, Sonnet 5.5)
- [GitHub integration]
- Mapping between claude.ai/projects and local folders limited to 5 projects
- ECONNRESET persistente (attempt 10/10) — RST originado por la IP de la API, no por la red local
- [GitHub integration]
- [GitHub integration]
- [FEATURE] Allow rebinding Ctrl+D to scroll:halfPageDown in fullscreen (vim-style scrolling)
- Projects (beta) feedback
- [BUG] Sessions repeatedly auto-archived during active use, no way to unarchive
- [FEATURE] Slides artifact type with presentation mode, speaker notes and team collaboration
- [Bug] /goal stop check rejects a user-ordered handoff repeatedly (about 30 times in a row), with no bound and no way to pause
- Auto-mode classifier denies the Telegram plugin's own reply tool as "Unrequested Commit in a Connected App" when answering an inbound Telegram message
- [GitHub integration]
- fix(plugin): strict validation skips malformed skill and agent files
- reasoning_extraction rejection on ordinary configuration request after visible history ingestion
- [Desktop] Split view: "+ New", Ctrl+N, deep links and pop-out dock all overwrite the leftmost pane; no way to open a new session in a new pane
- [GitHub integration]
- Custom subagent frontmatter model (haiku/sonnet) ignored — 1,149 of 1,371 subagents ran on parent Opus model (2.1.277–2.1.290, intermittent)
- [BUG] $.model.fork misses the conversation prompt cache when Message Threads and a mid-conversation system message are both active (2.1.291)
- [BUG] Windows desktop app: startup helper lists a remembered Code-tab folder on a Google Drive virtual drive, hangs unkillably, and blocks the app from relaunching
- [Feature Request] Enable standard copy/paste keyboard shortcuts for space navigation
- [BUG] Desktop (macOS): main process repeatedly navigates the Code tab to nonexistent /epitaxy/n9, replacing the window with "Page not found"
- message flagged - no reason
- [BUG] Auto-update removes the macOS Desktop (Space) assignment for Claude
- [BUG] Desktop (Windows): Code tab sidebar collapses during computer use and stays collapsed in a maximized window
- Desktop Code tab: bring back the live token count (now only shows a timer)
- [Feature Request] Add option to disable classifier
- [BUG] Desktop: project thread's PR bar shows Auto-fix CI unchecked while its Remote Control session's local monitor is on
- [BUG] GitHub App shows "Configured" but repository not appearing in repo picker on claude.ai/code
Resolved issues (156)
- [GitHub integration]
- [GitHub integration]
- [Feature Request] Add authentication/verification for shared deobfuscation workflows
- Desktop app: allow custom styling of chat message bubbles (e.g. user-stylesheet or theme colors)
- [Bug] Account flagged for suspicious activity despite Cyber Verification Program enrolment
- [GitHub integration] Repo linked via "Add from GitHub" in chat, but Claude gets no GitHub tools and cannot read it
- [False Positive] Supported Countries Policy suspension on a US-based paid account
- [BUG] [BUG] Windows: OneDrive-redirected Desktop folders appear as duplicates in folder picker, require adding twice
- [BUG] Sidebar mixes project and standalone chats with no project label after Projects upgrade (follow-up to #87346)
- chat cant access private repo[GitHub integration]
- dd[GitHub integration]
- pipe
- GitHub option greyed out in Project knowledge despite connected GitHub account
- [BUG] Downloads from an HTML file opened in a conversation: percent-encoded `data:` links fail with an atob InvalidCharacterError; base64 links work
- Clause is freezing when trying to export a project to pdf file
- [Bug] Agent workflow performance degradation on trivial tasks
- [Bug] Excessive request blocking or rate limiting preventing normal operation
- [Bug] Overly aggressive safety filtering causing false positives on unrelated content
- [Bug] Agent processes hang on completion and require manual termination
- [GitHub integration]
- I'm ready to help generate GitHub issue titles for Claude Code bug reports. Please provide the bug report details, and I'll create a concise, technical issue title following the format specified.
- [GitHub integration]
- I appreciate you reaching out, but I need to respond in English since this is a technical system for generating GitHub issue titles in English.If you have a bug report or feature request for Claude Code, please describe the issue in English and I'll help
- プロジェクトでローカルのアプリをcompute useできない
- [GitHub integration]
- [Bug] Anthropic API Error: Authorization or Access Blocked for Cybersecurity Use Case
- [GitHub integration]
- Subagent iTerm pane title shows agent *type*, not the assigned --name — hard to tell parallel panes apart
- Interrupted Agent (subagent) dispatches return no partial-progress report — parent cannot distinguish "never ran" from "ran 19 minutes, stopped mid-action"
- Custom MCP connectors don't appear in /schedule routine-attachment connector list despite showing "Connected" in Settings
- Memory reads silently fail when Claude-3p workspace sandbox excludes ~/.claude directory
- [BUG] Windows: Cowork local sessions fail with spawn ENAMETOOLONG once ~400 scheduled tasks exist (--allowedTools exceeds 32,767-char command line limit)
- [BUG] All claude.ai connectors stopped attaching on 2026-07-28: no claude.ai section in /mcp, no proxy transport log written since
- Sidebar: add "group by environment" (show Local + Cloud/SSH/WSL sessions together, separated)
- [BUG] Bash tool fails on every command: "export: not valid in this context: SENTRY-TRACE" (zsh) — recurrence of #70416
- Scheduled task runs intermittently ignore the queued task prompt and greet instead of executing
- Read tool: `pages=` reroutes PDF reading through pdftoppm (hard-fails without poppler, 100-DPI JPEGs with it) while plain Read is fully native — and the tool description steers models onto the failing path
- Background agents can exhaust the machine unnoticed: no visibility into spawned processes or load
- effort 'max' is selectable while thinking is disabled, then fails with API 400
- Connector URL is read-only: mis-registered MCP connector cannot be fixed
- Kaggle MCP OAuth fails at token exchange: "client_secret_basic authentication requires a client_secret"
- Bash tool spawns child processes with no Win32 console on Windows — native console apps hang indefinitely
- Bedrock: x-anthropic-billing-header and system prompt injected into user message body when context is large
- Model destroyed uncommitted work with directory-scoped git restore while cleaning up another agent's rm accident — after stating the bulk-op lesson itself
- VSCode extension: pending dialog silently lost when a scaffolded folder is opened in the same window
- [Bug] Cannot resume subagent with CLAUDE_CODE_DISABLE_BACKGROUND_TASKS=1 and CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=0
- remote-control server spawns sessions via its resolved install path — every auto-update breaks new-session spawns with ENOENT until the server is restarted
- Interactive REPL: every message fails instantly with 'Unable to connect to API (ConnectionRefused)' — non-interactive paths work fine
- [BUG] Thinking block summaries truncated mid-sentence (showThinkingSummaries: true)
- [Bug] Plan sharing between forked conversations without warning causes unintended side effects
- [BUG] Desktop global search doesn't match renamed/pinned session titles
- Subagents are resumable conversations but cannot be resumed on a different model — model-scoped failures force a full restart
- [BUG] Background Bash task launched by a subagent is killed when the subagent returns; the promised completion notification can never fire
- [BUG] Registering a DirectoryAdded hook makes one scaffold Task subagent return no output even though the hook never fires
- [Bug] Bash commands executed in accept-edits mode without confirmation
- A tool listed in permissions.allow is denied by the permission classifier
- Desktop-to-mobile session sync is stale/missing for some conversations
- Cowork: Chrome "Approved sites" domain approval resets every new conversation despite choosing "Always allow"
- Sessions started from an assistant-suggested background task have no parent linkage, so findings get misrouted
- Custom keybindings.json not respected in Claude Code Desktop (Windows)
- Edit tool calls intermittently rejected even after adding a matching permissions.allow rule
- Desktop app session list omits sessions created by the VS Code extension or terminal CLI (private index vs shared transcript store)
- Fuzzy matching for # Slack-channel autocomplete in the CLI composer
- Sidebar project grouping is inconsistent; add a way to disable directory-based grouping
- [BUG] The harness dirties every isolated worktree via .claude/settings.local.json, which permanently disqualifies it from auto-cleanup — a read-only agent still leaks a full checkout
- [BUG] Searching within a rendered markdown file switches the viewer to code mode, and returning to rendered view resets scroll to the top
- [BUG] Desktop file viewer serves a stale cached copy of an already-open file; edits only appear after navigating away and back
- [BUG] Desktop app sessions are declared "non-interactive" (isTTY heuristic) in the MCP-auth notice, misrouting OAuth remediation
- PreToolUse hook additionalContext loses trailing newline on history rebuild, breaking prompt cache at the first tool call of every turn
- Surface MCP tool annotations (readOnlyHint/destructiveHint) in PreToolUse hook input
- Typing over a double-click-selected word inserts instead of replacing (prompt editor)
- Windows desktop app 1.24012.11.0: updater removes its own MSIX package out from under the running app, then reinstalls the same version in a loop
- Ambiguous click semantics in permission-prompt UI causes accidental wrong selections
- Ctrl+B does not background the running command until the next keypress arrives (2.1.221)
- [BUG] Terminal tab title never updates when a custom agent is pinned (--agent / settings.agent) — agentTitle suppresses topic-title generation
- VS Code extension: clickable file links don't open files with square brackets in path (e.g. Next.js [locale] folders)
- IDE extension hangs indefinitely when the workspace contains more than one folder (multi-root)
- [Bug] SendMessage tool shutdown_response reason field lacks rejection-only constraint documentation
- Scheduled tasks cannot pre-authorize browser-tool permissions - unattended browser automation stalls on per-action prompts
- Cowork (Windows): plan9 share of a mapped SMB network-drive folder goes stale mid-session and never remounts; sessions end without outputs
- MCP connectors (claude.ai / Claude Desktop) never fall back to anonymous access when a server advertises optional OAuth
- stdio MCP server: tools are not re-registered after an automatic respawn — one call succeeds, then "No such tool available" for the rest of the session
- Surface cumulative re-read cost in /context: 97.7% of tokens read across 608 sessions were re-reads, and 60% of sessions account for 93% of it
- Interactive sessions background themselves by default instead of offering it
- claude CLI 2.1.215 emits tool_use with empty input in stream-json (tools never execute)
- Claude in Chrome: extension never registers with cloud-hosted Cowork sessions, while local chats connect fine
- Bash tool rm permanently deletes files (bypasses Recycle Bin), no built-in safeguard against agent-initiated data loss
- [Bug] Claude Code performs write operations outside scoped file in plan mode
- Long-running background Bash tasks are SIGTERMed (exit 143) at unpredictable times during long interactive sessions
- test-hook.sh reports success for a hook that allows what it was written to block
- [BUG] Terminal mode: messages enqueued after first turn are never dispatched (CLI-side queue stall, no exthost restart) — v2.1.221, macOS 26.6
- [BUG] Desktop: integrated terminal never binds in a restored long-lived session; each click leaks a shell
- The coordinator cannot access my pc locally. He gives me prompts to copy like i…
- [Bug] Auto-classifier executes block action without providing reason
- `claude plugin eval` refuses every Bash-granting run on a Mac with Docker Desktop (symlinks in ~/.docker/cli-plugins)
- [GitHub integration] Github x Claude auth Loop
- /model picker: pressing Right on the highest effort level wraps to Low and saves it as the default
- [Bug] Incomplete Bug Report: Missing Details
- [GitHub integration]
- [Bug] Mouse click on plugin Button never fires onPress in fullscreen TUI
- [BUG] 2.1.286 idle compaction silently discards working context in long-running sessions; no opt-out, logged as "manual"
- [FEATURE] Add Russian (ru-RU) to the shipped UI locale set
- Projects (beta) feedback
- EnterWorktree fails from a subagent with a pinned cwd (isolation/explicit cwd), with no documented fallback
- PushNotification reports "Remote Control inactive" in sessions started with claude rc
- MCP tool results silently replaced by an unreadable <<ccr:...,html,NNNB>> placeholder (two unrelated servers)
- Hosted connector drops `blob` from embedded resources: -32602 invalid_union for any binary document (same payload accepted by the engine client)
- [GitHub integration]
- Muhasebe
- gladius
- [GitHub integration]
- [BUG] Predictive/ghost text appearing in input box with full solution
- [Desktop/Windows] Main webview renderer crashes on launch (0xC0000005) since 2.19675.0, app unusable
- [FEATURE] Let a published Design System theme the new-chat usage stats card (opt-in, token-based)
- [Bug] Code-review feature generating false positive distillation warnings
- [Bug] Actions execute without confirmation in manual mode
- [Bug] Agent status remains active after task completion with no running tasks
- [Bug] Safety classifier blocking legitimate operations during analysis and UI rendering
- [Bug] Claude code misclassifying QA test conversations as cybersecurity content
- [Bug] Missing environment and device status validation on session restart
- Overly sensitive security tripwire
- [Bug] Model ignores explicit tenant isolation constraints in database queries
- [Bug] Model generating unnecessary changes and poor decision-making during code modifications
- [Bug] Safety guardrails blocking legitimate game development use cases
- Limpieza antes
- [GitHub integration]
- USELSS PIECE OF SHIT CANT EVEN DOA. SIMPLE AUDIT
- The folders are not connecting and after reset the system has used 5% of my usa…
- Claude Error, requests not going through
- [GitHub integration] wind cloud
- [Bug] Unclear issue description - please provide bug report details
- [GitHub integration]
- فولدر المشروع مش راضي يشتغل علي كلود مع انه معمول ليه ريموت كنترول
- [GitHub integration]
- Marcia
- [Bug] Anthropic API 529 Overloaded causes subagent termination without retry; permission classifier outage blocks all tool execution
- Deferred tool schemas are not restored after context compaction, causing repeated InputValidationError
- [Bug] Summary text obscured by background task completion banner
- 2.1.222: worktree-isolated session refuses any command containing an unresolvable env var (echo "$TERM" is blocked)
- Persistent Monitor tasks killed with SIGURG (exit 144) ~60-75 min after arming; trap '' URG cannot prevent it
- /simplify review fork escalated from report-only scope to committing, submitting, and merging a PR
- [BUG] Worktree cleanup is unsafe on Windows — 'git worktree remove --force' destroys NTFS junction targets, so vendor-created worktrees accumulate with no safe way to reclaim them (130 GB, ~30% of a week's usage spent trying)
- [BUG] Grep honors .gitignore silently — an agent auditing its own repo gets 'No files found' for code that exists, and cannot tell blindness from absence
- Async subagent killed by a mid-stream API error still reports task status "completed", with the error text as its summary
- Compaction fails with 400: messages.N.content.M.text.parsed_output: Extra inputs are not permitted
- Esc key semantics are inconsistent and make stopping an agent dangerously easy (CLI)
- Stop hook exit-2 verdict silently discarded (no stop_hook_summary logged) when turn ends on a tool result with a pending ScheduleWakeup
- Auto-compact fails at context limit and forces a new session, while manual /compact succeeds
- MCP OAuth: org connector does not try path-based well-known discovery for path-prefixed servers
- Stats "Peak hour" ignores the 7d/30d/All range selector — hourCounts is a flat dateless histogram, and it goes stale
- [BUG] design-sync truncates prop JSDoc mid-word at 120 chars with no marker (lib/dts.mjs:434)
- Agent view row shows the parent/default model, not the per-spawn `model` override — same root cause you already fixed for `/agents` in #20291 (v2.1.199)
- [BUG] JetBrains plugin: "Send to Claude Code" (Ctrl+Alt+K) does nothing when the Markdown preview pane is focused
- MCP tool call issued during session re-initialisation is silently discarded
- Session cwd silently resets to original launch directory, discarding legitimate navigation (affects PreToolUse hook cwd for all tool types)
- message flagged - no reason