Digest / September 5, 2026
Claude Code Issues Digest — September 5, 2026
162 new issues filed ·
657 resolved in anthropics/claude-code (UTC day).
New issues (162)
- [BUG] Agent teams: one teammate message or idle_notification is delivered to the lead up to 4 times with an identical timestamp (successor to #74112, stale-closed)
- [Bug] Anthropic API Error: Safety guidelines blocking legitimate game development requests
- Windows desktop app self-updates and restarts over a running session — nine forced restarts in nine days, no opt-out
- [MODEL] CLAUDE 5 CLI SAID SONNET 4.6 NO LONGER AVAILABLE
- Browser pane tool fails on all read/screenshot calls with "Policy check temporarily unavailable"
- [BUG] ListAgents / SendMessage missing from tool registry in Desktop scheduled-task and Remote Control sessions (bisected to Desktop 1.44121.4 -> 1.46388.1)
- [Bug] Auto mode requires manual input instead of operating autonomously
- Scheduled task (Cowork) tự động fire trùng lặp nhiều lần trước giờ đã lên lịch — tái diễn 5 ngày liên tiếp
- CLAUDE_CONFIG_DIR set to the default path ($HOME/.claude) selects a different Keychain credential entry than leaving it unset, with no diagnosis
- [Bug] Frequent model reset required during interactive sessions causes workflow disruption
- [DOCS] `--setting-sources` docs do not say what each source value actually gates (CLAUDE.md, plugins, hooks, skills, agents, commands, `--settings`)
- [BUG] MCP tool schemas still consuming context tokens after all connectors disabled
- [Bug] Claude Code commits changes without confirmation when explicitly instructed not to
- Re-inject user-marked critical rules before each turn — gating rules decay between turns while format rules hold
- SendMessage/ListAgents cross-session peer messaging broken after CCD 2.1.258 -> 2.1.260 update (Windows desktop app)
- [FEATURE] Agent(agent_type) allowlist in tools has no effect inside a nested subagent definition
- Desktop (Windows): claude://code/new?folder= adds the folder to the open workspace instead of opening a new window, re-prompting for trust every time
- All models behave the same way repeatably
- Bash tool: heredoc file writes deadlock and truncate the target file on macOS with Homebrew bash
- [Bug] Anthropic API Error: Content flagged by safeguards on benign input
- Idle background session stops advancing while async subagents are in flight, and resumes only when the operator opens it
- [BUG] VS Code extension: Remote Control session shows "Connected" but never consumes messages from mobile/web; extension processes accumulate
- [BUG] Claude Code desktop: closed sessions leave `claude` child processes running — possible regression of #58915/#61748
- [Bug] Model creates Jira tickets without explicit affirmative approval despite standing instructions
- [BUG] Cowork: agent-created scheduled tasks can never be device-bound, and "Require this computer" cannot be enabled afterwards
- [BUG] Desktop app: custom sidebar groups render empty while claude.ai shows the same groups populated
- [Bug] Claude Code makes incorrect assumptions about user intent without confirmation
- bypassPermissions Bash-first instruction (thrifty_sonic) silently skips path-scoped rules, nested CLAUDE.md, and Read/Edit/Write hooks
- [Bug] Task Output lookup fails when reports arrive after parent recovery from subagent files
- Cowork sandbox network allowlist not affected by sandbox.enabled toggle in settings
- stable channel has not advanced past 2.1.236 since Aug 19, against a documented cadence of about one week
- Terminal identification queries are sent during a ~50ms window where ECHO is re-enabled, leaking replies to the screen at startup
- [BUG] Desktop 1.44121.4+ never auto-enables Remote Control for scheduled-task sessions (regression from 1.40609.0)
- Claude refused direct instruction to include user-provided content in a draft
- [BUG] Regression since 2.1.259 on macOS: EACCES traversing /Library/Managed Preferences is treated as an unreadable managed policy and fails closed (works in 2.1.258)
- Auto mode: let a classifier block fall back to a permission prompt instead of a hard deny
- Marketplace from a local git directory installs the checked-out branch, and autoUpdate does not keep the cache current
- [Bug] Model lacks confidence in derived answers, causing inefficient loops
- [Feature Request] Add list creation and collaborative task review workflow
- Background task kill/completion notification wakes the session and re-reads full context -- costs a full turn with no user input
- [BUG] GitHub Connector — "state: Field required" error blocks repo access
- n8n Vue Flow canvas renders blank (viewport transform stuck at "none") when driven via Claude in Chrome
- [BUG] Assistant generates fabricated "user" and "system" lines in its own output (17 cases, 5 sessions)
- [FEATURE] NVDA and not able to select text in the prompt box.
- [FEATURE] Local status feed from the desktop app (Cowork, Chat) for ambient "Claude needs you" tools
- [BUG] Claude Desktop blocks CLI-native SendMessage, breaking subagent continuation
- [Bug] Subagents inherit session model instead of respecting explicit model specification
- [Bug] No notification when model automatically downgraded due to safeguards
- [BUG] Desktop app: file tree shows a directory symlink as a file and cannot open files under it (regression)
- Self-report: autonomous session overrode an explicit real-money instruction, added unrequested features, deployed without viewing UI (claude-fable-5-1, 2026-09-05)
- [BUG] show_widget: failed validation call + retry with same title leaves duplicate widget rendered
- [FEATURE] Desktop: Ctrl+Tab should cycle the working set of sessions, not every session in every project
- [BUG] Scheduled-task sessions stopped auto-connecting to Remote Control on 3 Sept — the "Connect new sessions" preference does not cover them
- [Bug] Main session violates established procedures repeatedly
- Worktree/branch auto-naming gets stuck on one name ("spreadsheet-sync") across unrelated sessions in the same repo
- [FEATURE] Add an "auto-daltonized" theme mode (easy change)
- [Bug] Claude Fable 5/5.1 model not available despite within org limits
- execute_zapier_write_action / execute_zapier_read_action fail with "Invalid input: expected record, received string" (path: params)
- [Feature Request] Add granular subagent topology control, execution monitoring, and drift detection
- [BUG] Claude Desktop (macOS) still bundles darwin-x64-only node-pty, triggers Rosetta/Intel-app deprecation warning
- [FEATURE] LSP: filename-based routing (filenameToLanguage) for extensionless canonical files such as justfile, Makefile, Dockerfile
- [BUG] `claude -p` with piped stdin counts the input ~1.78× toward the context limit ("attachment content"), halving the usable window for large prompts
- Desktop app: interactive sessions never register for Remote Control, while scheduled-task sessions from the same app do
- [BUG] Claude Code 2.1.261 crashes
- [Feature Request] Email notifications for scheduled routine triggers
- [Bug] Claude ignores user-specified requirements and reverts to discarded features
- [Bug] Repeated "Policy Violation" errors on all tasks
- [Bug] CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS is skipped when ultracode is active — a Workflow ran 7 agents under a cap of 3 (2.1.261)
- [BUG] Linux: Alt+V image paste does nothing — chat:imagePaste is bound to Alt+V only on Windows/WSL, so default shortcuts differ per OS
- Desktop app: sessions silently drop out of their sidebar group while in use
- [DOCS] Desktop Code tab disables the CLI-native SendMessage, but the docs say cross-session messaging reaches terminal sessions
- Feature request: create a new session directly in a split pane (one-step)
- [BUG] API Error: 400 invalid params, messages.4.content.1.tool_use.input: Input should be a valid dictionary (2013)
- [BUG] claude-security scan-changes reproducibly analyzes wrong diff content
- Restarting Claude Desktop on two machines registers each device twice; sessions then leave their project group and appear under "Other" on the other machine
- [Bug] Auto-resume after usage-limit reset consumes session time silently
- [Bug] Claude ignores explicit instructions to show output in chat instead of writing to file
- Remote Control bridges die on desktop auto-update relaunch; rcAutoEnable only re-arms on a local turn
- [BUG] MCP connection to Wolfram
- [BUG] CoworkVMService kann seine SCM-Wiederherstellungsaktionen nicht konfigurieren
- [Bug] Claude Code ignores user guidance in memory/rules and repeats mistakes independently
- [BUG] Desktop: stalls over 5s are discarded as "[likely sleep: duration_heuristic]" and dropped from the app's own counters — 91/1129 samples, zero sleep events in Windows' power log
- /skill-doctor's per-skill "7d tokens" / "uses" numbers don't reflect a skill's actual cost or invocation count
- /skill-doctor's per-skill "7d tokens" / "uses" numbers don't reflect a skill's actual cost or invocation count
- [BUG] Write emits .cmd as UTF-8 + LF-only on Windows; cmd.exe (CP932) swallows the LF after a multibyte char, del /F /Q "%VAR%" degenerates to del /F /Q "" and deletes every file in the CWD
- [Bug] Cybersecurity filter incorrectly flags password management system creation
- Remote Control auto-start reports "re-login required" when the OAuth refresh simply hasn't landed yet (credential still renewable)
- [FEATURE] Let claude.ai chat delegate briefs to a Claude Code session through an approval card, and get the result back
- [Bug] Excessive agentic loops for simple code edits
- code-review skill coordinator freezes/fabricates status even after all background subagents finish
- [Feature Request] Support for low-cost microcontroller platforms (ESP32) in educational workflows
- [BUG] Claude in Chrome 1.0.91: Sign-in screen after every browser-session end (reboot / profile switch) — silent re-auth never restores the session while claude.ai stays logged in
- [Bug] Excessive token consumption during repository scanning
- Windows: main window intermittently gains WS_EX_TOPMOST (breaks Alt+Tab) — captured correlation with LocalSessions.setFocusedSession
- [BUG] Stop hook reports phantom "unpushed commit(s)" after a squash-merged PR whose branch was auto-deleted
- Web UI: message sent after dismissing an AskUserQuestion popup renders out of chronological order
- [Bug] Claude Code lacks reasoning for multi-step function implementations
- [Feature Request] Add session management controls: immediate deletion, pinning, and TUI dashboard
- Feature request: a hook that fires after a session releases its worktree, so cleanup can succeed
- [BUG]
- [Bug] Windows CMD terminal opens duplicate tabs when clicking links
- [BUG] Stray priconfig.xml in Desktop MSIX breaks installation with 0x80073CF9
- System-reminder wrapper for CLAUDE.md/MEMORY.md uses injection-shaped language ("OVERRIDE any default behavior... MUST follow exactly as written")
- [Bug] "Summarize up to here" silently drops earlier compact summaries located after the anchor - permanent context loss
- The prompt suggestion should steer toward the right next action, not the most predictable one
- [BUG] Ultracode removes proportionality: "use the Workflow tool on every substantive task; token cost is not a constraint" turns a settings audit into 325 agents and a design task into 16, re-reading what the main loop already read, on a sticky effort setting the user cannot see
- [BUG] /permissions silently drops existing rules with lowercase tool names when adding a new rule
- Password-handling guardrail applies uniformly, ignoring low-stakes dev context, and can contradict itself
- [BUG] Closed Weighted Training Overrides Rules
- [BUG] UserPromptSubmit hook registered and valid but never invoked on real prompt submission (Git Bash CLI, Windows, v2.1.261) — possible regression of #17277
- Plugin cache: a new version dir inherits the previous one's untracked files, replicating them forever
- I appreciate you reaching out, but this message doesn't contain a specific technical bug report or feature request that I can convert into a GitHub issue title. To help address your concerns effectively, please provide: 1. **Specific problem**: What exac
- [Feature Request] Add conciseness mode or response brevity control to reduce verbose output
- Scheduled tasks: expose a per-task directory-access grant (parity with local-agent-mode's userSelectedFolders)
- [BUG]
- [BUG]
- Voice dictation permanently freezes TUI: main thread blocked in synchronous CoreAudio call (AudioUnitSetProperty) while opening microphone
- [Bug] Claude prioritizes tool execution even though it is explicit in the acknowledgment that the tools cannot be trusted.
- [self-built FEATURE - is it future proof?]
- [BUG] Rejected turn leaves entire session stuck in refusal loop, even for trivial prompts like "hola"
- [Bug] Agent ignoring memory context and performing unintended actions beyond scope
- A non-string entry in sandbox.filesystem.denyRead/denyWrite silently disables permissions.deny and PreToolUse hooks (2.1.261)
- Voice dictation should keep recording through permission-approval and diff-review confirmation dialogs
- [Bug] Anthropic API Error: Incorrect Safeguard Flag on Non-Cybersecurity Content
- [BUG] Opening text of the turn after a tool-call rejection is dropped from display AND the session transcript (model/user conversation diverges)
- [BUG] Session resumes and executes further tool calls after a hard trust-safety block, via background-task completion notification
- Claude in Chrome: all sites blocked by site permissions, persists across full reinstall
- [Bug] Session expires too quickly with low priority requests
- Feature request: opt-in local bridge so scheduled cloud agents (routines) can use local MCP servers
- [Bug] Anthropic API Error: Model switching to Opus 5 causing degraded code quality
- Chat list ordering destroyed by batch mtime restamping of old transcripts
- [Bug] False Positive Safety Flag on Git Workflow Discussion
- [Bug] Tool execution consumes token budget unexpectedly fast
- [BUG] --name is not applied to the peer registry on 2.1.261; new sessions register under another session's generated title
- [BUG]
- CoworkVMService fails to configure SCM recovery actions — "Access is denied" on every start
- Add setting to make active-file/selection context opt-in (manual) instead of automatic
- [Bug] Claude Code incorrectly flags legitimate queries with Fable 5 safeguard warnings
- Desktop app: per-chat or per-group background colour so the chat pane shows which project it belongs to
- [BUG] Cross-session send_message from an auto-permission-mode sender is silently dropped (45s no-ack settle); bypassPermissions sender delivers
- Recurring spurious [bio] refusal (empty content) across unrelated sessions, same day
- [Bug] Safety filter false positive on test failure analysis requests
- Unrelated <system-reminder> injected into Bash tool result, instructing to add Claude attribution to commits
- xhigh effort level no longer available on Opus 4.6
- I need more information to generate a GitHub issue title. Could you please provide: 1. What specific problem or error are you experiencing? 2. What were you trying to do when it happened? 3. Any error messages or unexpected behavior? 4. Which version of C
- [BUG] theme: "auto" never updates on Windows Terminal — resample is gated solely on DECSET 2031, which WT does not implement
- Thanks for the extra token budget!
- [Bug] False-positive bio safety filter on materials-science EIS tortuosity task
- [BUG] Went from 0% to 100% U5 hour usage in 10 min on Opus 4.8
- [BUG] 2.1.261 macOS arm64: session hangs, main thread 100% CPU in JIT loop while waiting on a background subagent (sample attached)
- permissions.defaultMode is ignored by the desktop app - new sessions always start in manual mode
- [BUG] Desktop: Terminal panel flickers open/closed on every shell tool call, stealing keyboard focus (Windows)
- [Bug] Incomplete system integration affecting core functionality
- [FEATURE]
- Permission-denied message shows a sub-command excerpt of the Bash command, not the full command or the matching deny rule
- `run_in_background` Bash call returns "completed" immediately when the script backgrounds its own children, and a follow-up `ps` can miss the still-running children
- [FEATURE] @-file picker: Tab should complete to the longest common prefix, not full-accept the top match
- Auto-update regression (2.1.255) broke launch on macOS Monterey with no warning or rollback path
- [FEATURE] Keyboard shortcut for microphone in main chat window (Claude Desktop macOS)
- [BUG] Drag & drop from the VS Code Explorer silently does nothing in Remote-WSL/SSH/Dev Containers — drop handler only accepts `file://` URIs
- / bug report
- SendUserFile: 送信した画像に常時表示のダウンロードアイコンが出なくなった
Resolved issues (657)
- [Feature Request] Add guardrail exemption or allowlist for trusted repositories
- bypassPermissions Bash-first instruction (thrifty_sonic) silently skips path-scoped rules, nested CLAUDE.md, and Read/Edit/Write hooks
- Desktop app (macOS): AskUserQuestion question body renders without line breaks
- [BUG] Assunto: Claude Desktop (Windows MSIX) trava repetidamente e corrompe o pacote — persiste após reinstalação completa do Windows
- I don't have a bug report to analyze. Please provide the bug report or issue description you'd like me to generate a GitHub issue title for.
- I appreciate you sharing details about your authorized bug bounty work, but I need to clarify something important: **This is not a bug report for Claude Code.** You're describing security testing work you're conducting on a third-party e-commerce platform
- [BUG] Dispatch missing from Claude Desktop on Windows — visible on mobile (same account), pairing always fails (related: #82431)
- I appreciate your interest, but this isn't a bug report or feature request for Claude Code. This appears to be a job inquiry. If you have a genuine bug to report or feature to suggest for Claude Code, please provide details about: - What problem you encou
- [Bug] Unrelated safety filter blocking legitimate Bluetooth operations
- [Bug] Model incorrectly shifts to bio task during coding operations
- [Bug] CTF Challenge Flagged as Suspicious - Legitimacy Verification Needed
- [Bug] Unclear error message when request is stopped
- Planning-instance failure pattern: abandons committed lane mid-arc, then cites unverified self-built checkers as evidence (claude.ai, Fable 5)
- Opus 4.8: two consecutive 64K-token runaway thinking generations with zero visible output (empty thinking blocks)
- [BUG]
- [BUG] Remote Control presence pulses get intermittent 401 on healthy connected bridges (no retry) — session shows offline/missing in mobile session list
- [Bug] Excessive resource consumption - tasks using 20x normal resources
- [BUG] Remote Control intermittently fails with 'Remote credentials fetch failed' / 'Session creation failed': transient 401s from code-session service exhaust the 3-attempt retry budget
- Scope overstep on a GUI-launch task: model mutates persistent window/launch state it was never asked to touch, and repeats it across corrections (claude-code, Opus 4.8)
- [BUG] Cowork on Debian 12: helper resolves system C virtiofsd instead of bundled Rust virtiofsd, causing setgroups/FUSE failures
- Workflow({name}) does not discover saved workflows under .claude/workflows/, contradicting the tool description
- [Bug] Anthropic API Error: 529 Overloaded
- Model executes a destructive persistent state change (deletes OS-level saved settings) inferred from a goal-level instruction — no confirmation, no backup, Nth same-session recurrence
- Claude in Chrome: Page.captureScreenshot times out (30s) on any tab whose visibilityState is "hidden" (renderer backgrounded while the desktop/chat window is focused); same mechanism as #53475
- Fable restrictions
- [BUG] Custom connector tools never reach new conversations since Desktop v1.17377.1 (directory connectors fine)
- [BUG] Server-side web_search blocks from an LLM gateway are never rendered — no "Web Search" line, and citation-segmented answers fragment into one bullet per block
- [BUG] GitHub repo-scoping gateway returns 403 on github.com even with "All domains" egress, breaking Bazel dependency resolution
- MCP stdio server fails to (re)connect with error -32000 after update to 2.1.212, though the subprocess handshakes correctly when invoked directly
- [Bug] Misleading tool capability claims and unverified verification assertions
- [BUG] Claude Desktop code agent drops MCP per-tool `_meta` — `anthropic/maxResultSizeChars` ignored (CLI honors it)
- PreToolUse hook's empty stdout forces an interactive prompt, ignoring an existing allow rule (VS Code extension)
- Anthropic: Go fuck yourself with your safeguard!!!!!!!!
- Anthropic: Go fuck yourself with your safeguard!!
- Auto-updater (ShipIt) doesn't relaunch the desktop app after installing an update — breaks anything relying on the app staying open
- [Bug] Plugin hooks silently dropped when plugin has no version
- Session limit still enforced after plan upgrade — upgrade doesn't lift the cap it was purchased to remove
- [Bug] Anthropic API Error: Model access denied for claude-fable-5
- Cloud sessions (`/ultraplan` handoff) die on first turn with generic "An error occurred while executing Claude Code" when repo-tracked settings enable sandbox with `failIfUnavailable: true`
- Gmail connector omits Inbox messages classified as "suspected spam" (delivered via Workspace admin override)
- [BUG] Backend claude process and its MCP/socat children are not terminated on tab close, session resume, or project deletion — leading to unbounded process accumulation.
- [BUG] [Cowork] New project's local folder context fails to load - mcp__remote-devices__device_list_dir not registered in session
- Claude in Chrome: Desktop wins host-selection race by default; forcing Claude Code's own host to launch (manifest-disable + extension toggle) still leaves MCP handshake incomplete
- [BUG] Cowork auto-start on launch wipes the entire `mcpServers` block from `claude_desktop_config.json` (Windows MSIX)
- [BUG] [claude-in-chrome] Typed text mis-inserts into the document body during the Google Docs comment workflow (focus/state race + mid-task connection drops)
- Native-install auto-update rotates the versioned binary path, invalidating macOS TCC grants attributed to Claude Code (Calendar/Reminders observed)
- [BUG] Claude-in-Chrome via MCP: per-domain access blocked, not synced with side-panel "approved sites"
- [BUG] Cowork "Progress" panel shows false "tool not found" for Desktop Commander (Windows, after app update)
- Gmail MCP search_threads returns stale/incomplete results for long-running threads; get_thread by ID is accurate
- [BUG]
- [BUG] SessionStart hook stdout not injected into context (regression 2.1.209 -> 2.1.215, desktop app)
- Windows: stdio MCP server spawning ssh.exe never connects (same server works via macOS client, Codex CLI, and manual pipe)
- [BUG] Scroll Lock key causes Claude Code to exit in VS Code integrated terminal (Windows 11)
- [Bug] Cron jobs failing due to stale server references and stopped service dependencies
- [BUG] Desktop "Open in terminal" says "session resumed" but creates a new session
- [FABLE] Persistent multi-week inability to use Fable 5 for device/sensor-heavy (camera, mic, external-fetch) feature work — likely dual-use safety classifier over-triggering
- FUCK off anthropic you shit fuck heads I AM WORKING ON MY...model
- [BUG] Claude in Chrome: phantom tabs accept and answer javascript_tool execution after Chrome has closed them
- [BUG]Safety classifier false-positive on formal verification / EDA work
- Built-in Bash sandbox: network isolation unreliable on WSL2; failIfUnavailable doesn't detect a non-enforcing proxy
- [BUG] /hooks omits hooks from managed-settings.json even though they execute
- Voice mode: captures good audio but transcription intermittently returns blank (WSL2/WSLg, v2.1.215)
- Routines: hourly cron minute is silently rewritten to the save-time minute (undocumented)
- PostToolUse async hook invocations partially dropped on consecutive Write/Edit batches (Windows, 2.1.139)
- [BUG] Claude Desktop 1.18286.0: tool-call cards render fully expanded and can no longer be collapsed (regression; bundled CC 2.1.197)
- TCC network-volume permission re-prompts on every update (binary path includes version number)
- Bug - Les instructions d'une routine creee via l'API /v1/code/triggers ne sont pas executees
- [Bug] Credit Usage Mismatch: Showing Depletion Below 50% Usage on max(20x) Plan
- [Bug] Plan mode restriction not applied after model switch with /model command
- [Bug] Incorrect usage quota calculation for MAX 5X tier with Claude 5
- LLM systematically terminates verification scope via L4 reasoning — defects missed
- Remote Control: reconnect burst wedges hosts in a no-backoff CPU spin (A); a busy turn is misread as "stopped responding mid-turn" (B)
- [Bug] Session rating registered on accidental number key press without confirmation
- [BUG] Claude Desktop Mac Rendering Issue
- [BUG] Outdated CLI (2.1.211) session hits spurious 'session limit' on Fable 5 and silently falls back to Opus 4.8 — no guidance to update/restart (works fine on 2.1.215)
- [Annoying] /login flow is not smooth
- Model selector reverts to Sonnet after "Fable 5 requires usage credits" dialog, despite Max plan including Fable 5 and `/model` reporting success
- [BUG] Remote Control Model Selector Defaults to EXTRA effort with no changeability, overrides local CLI
- [Bug] Question by the agent not visible in dark theme
- [BUG] Session silently stopped persisting to JSONL for 10 days while still working; resume after auto-update loads the stale state (Windows variant of #79298)
- [BUG] Remote-execution sandbox: LibreOffice cannot load any document — type detection fails for all formats, --convert-to exits 0 with no output
- Claude Desktop (Windows) — steady kernel nonpaged-pool memory leak tied to app runtime, causes full system hang after ~1.5-2 hours
- [Cowork] "Failed to load local file" is not txt/md-specific — `_readLocalHostFile` fails the same way for every file type, including xlsx/html which silently recover
- [BUG] Claude Chrome Extension - Cannot access certain sites
- [BUG] winget install succeeds but fails to create link — binary stranded in WinGet packages cache
- [Bug] Safety policies incorrectly blocking non-security/biology open source work
- Remote Control does not auto-reconnect after a network interruption
- Setting to hide/slim the input footer hints (mode indicator + "esc to interrupt")
- [BUG] Resource not accessible by integration" 403, GitHub
- [Bug] Onboarding guide sharing broken in 2.1.215: `check` crashes on e.length, `update`/`create` report success but write nothing
- [BUG] Cowork (macOS): Max→Pro downgrade wiped ALL session history from the UI — data loss of paid work, again
- VS Code extension hard-deletes conversation transcript on update/restart (no soft-delete, bypasses Recycle Bin)
- [BUG] macOS: locked login keychain silently breaks auth — /login falsely reports "Login successful", and entering the password at the keychain prompts still leaves the session logged out
- [Bug] Usage token activation required despite valid max tokens and available quota
- Fable 5 safeguards flag first-party defensive security audit, falls back to Opus 4.8
- Illegal instruction crash on ARM64 Linux in Apple Virtualization Framework (AVF) — embedded Bun runtime incompatible
- MCP tool calls with array-typed parameters (list[int]) arrive server-side as stringified text, failing validation
- [Bug] Fable model unavailable in CLI but works in desktop app
- [Bug] Anthropic API Error: Usage Policy violation false positives during security audits
- [Bug] Tool calls stall without execution, tokens consumed
- claude-sonnet-5 (Messages API, adaptive thinking high + tools): degenerate no-op bash loop instead of emitting long freeform output
- MCP server connects successfully but its tools are never exposed to the model
- [BUG]update_plan always returns "No tab available", blocking all browser navigation/screenshots
- --channels host mode uses stored Keychain OAuth credentials instead of valid CLAUDE_CODE_OAUTH_TOKEN (precedence inverted vs docs)
- [BUG] Recent sessions list shows stale/live-changing timestamps — items appear out of order or vanish entirely
- MCP plugin tools skip permission prompt when not in allow or deny list
- error 402 fake
- [BUG] Claude attempts to use unavailable `timeout` on mac
- [Bug] High token consumption and poor code quality with excessive tool use and compilation errors
- [BUG] Routine repository picker shows "no repository found" — GitHub connector authorized but GitHub App never installed
- [Bug] Anthropic API Error: Connection closed mid-response during streaming
- I need more information to create a specific GitHub issue title. Could you please provide: 1. What specific bug or problem are you experiencing? 2. What error messages are you seeing (if any)? 3. What behavior is occurring that shouldn't be, or what's mis
- Claude desktop app launches but shows no window on macOS 26
- Add confirmation before closing session with double Ctrl+C
- [BUG]
- Feature request: SubagentStart hook stdin should carry the Task tool description for subagent attribution
- [Bug] Fable 5 safeguards triggering false positives on literary criticism terms
- [DOCS] OpenTelemetry documentation does not explain CLAUDE_CODE_OTEL_CONTENT_MAX_LENGTH
- [DOCS] OpenTelemetry log-event reference omits new message correlation and tool provenance attributes
- [Bug] Anthropic API Safeguard False Positive on Academic Content in Markdown Editorial Review
- Browser pane: tabId rejected as missing despite documented "default: the fronted tab" (read_page, javascript_tool)
- [Bug] Anthropic API Error: False positive safety filter on PDF text extraction
- OpenTelemetry: bridged / remote-controlled sessions emit no local OTel metrics
- PreToolUse Bash hook not invoked for subagent (Task/Agent-tool) tool calls
- claude.ai connector MCP session silently wedges after ~15 min idle — no drop detected, no reconnect; fresh process succeeds on the same call simultaneously
- Headless -p mode hangs forever reusing a server-closed idle keep-alive connection after a long tool execution
- [Bug] Claude Code ignores CLAUDE.md prohibitions and pushes to protected branch without approval
- [Bug] Model switching from Fable to Opus causes hang/flag state
- [BUG] Claude kept saying "fixed" without checking — burned a huge amount of tokens over a month without actually fixing the bug
- [MODEL] Model presented unverified inference as fact, then used anthropomorphizing self-justification ("not being paranoid") instead of a direct correction when challenged
- [Bug] Anthropic API Error: Safety filters blocking request with valid CVP
- False "Context limit reached" on Sonnet 5 well below reported usage — not specific to skill loading
- Remote Control: no path to attach RC to completed headless (-p) sessions - resume requires a prompt, and -p --remote-control is silently ignored
- [Bug] Model not respecting claude.md instructions at runtime and leaking sensitive data in context
- [Bug] Claude 3.5 Opus model performance degradation
- [BUG] Claude code repeatedly presenting plan for approval
- [MODEL] Claude appended a string to its own reply, then treated it as user input on the next turn
- [Bug] Claude 3.5 Opus ignores explicit instructions and repeats programming actions without consent
- [MODEL] Opus 5: "verified" asserted from an unbounded search — recurs after correction, subagent inherits it, model launders its own errors into precedent
- [Bug] Document update blocked by Fable integration
- Trusted device re-registered every session instead of persisting, forcing frequent re-auth
- [Bug] All requests blocked by safety guardrails and tool execution overhead
- [MODEL] Opus 5 asserts an unverified repo-structure claim and defends it; when finally challenged it runs two commands and finds the inverse
- [Bug] Model ignores stored memory notes when selecting tools in subsequent actions
- [BUG]
- [Bug] Repository fork operation flagged as cyber risk
- [MODEL]
- Remote MCP servers declaring completions/logging capabilities fail to connect despite valid initialize handshake
- Drive de Google no conecta
- [Bug] CVP applicant hitting Anthropic API guardrails repeatedly
- [BUG] Native binary segfaults within ~2 seconds while nProtect GameGuard (Helldivers 2) is running — GameGuard module confirmed in crash stack
- [BUG] Thai combining-mark cursor bug has regressed (previously fixed in #22390) — v2.1.220
- [Bug] Context window limit reached during security-related conversations
- [Bug] Opus model not selectable in CLI
- [Bug] Incorrect working directory detection in file operations
- [MODEL] CLAUDE.md is systematically violated
- [Bug] Anthropic API Error: Safeguard flag causing unexpected model fallback during code explanation
- Arc: extension connects but all commands time out; /chrome always launches Chrome (hardcoded detection order)
- Domain-scoped credential injection for sandboxed sessions
- [Bug] Unexpected policy popup during extension/plugin development
- Per-session private /tmp, covering background jobs
- [Bug] Policy triggered during AI learning sketch creation
- [BUG] Wasted paid session quota by ignoring explicit constraints and looping on broken scripts
- Plugin cache never prunes superseded version dirs — 4.77 GB stale across 4 marketplaces (incl. first-party), 11 copies of one plugin, no cleanup command (2.1.220)
- [Bug] Content filtering blocks religious Arabic words across new chats
- [Bug] Fallback to Opus on routine coding tasks
- Remote planning session: deferred approval tool splits the turn, session drops out of plan mode ("not in plan mode"), plan is lost with no recovery path
- [Bug] Anthropic API Error: Request rejected (req_011CdPRch8H2R1YWazvtHBMF) - Potential misclassification
- After switching accounts, Keychain holds the new account's OAuth token but all identity surfaces still report the previous account
- [Bug] After ExitPlanMode and context clearing Bash tool is not loaded
- [Bug] Subagent permission prompt not displayed in mobile Claude app, causing task hang
- [Bug] Misleading documentation or error messaging regarding API usage or billing
- [Bug] Security filters blocking legitimate security audit requests on owned domain
- [BUG] sandbox.filesystem.denyRead does not block reads (file-type patterns and plain folder paths both fail)
- [Bug] Unexpected Model Downgrade to Claude 3.5 Opus
- [Bug] Claude ignores "do not push to production" instruction from claude.md
- Intermittent: subagents hang on their first tool call (0 tool uses, no error, parent waits forever)
- Model proceeds past explicit project prohibitions and scope gates that were present in context
- Model substitutes countable verification for judgment-requiring verification, then reports the work as complete
- [Bug] Fable 5 model unavailable in VSCode extension and CLI despite availability in Claude app
- [Bug] Authentication fails for test users without password enforcement
- [Bug] Fable-flag false positive during keystroke diagnostic instrumentation
- [Bug] Fable 5 rejects trivial arithmetic with contradictory error response (is_error: true, subtype: success)
- [BUG] Background Workflow dies at a session boundary regardless of the user's answer to the compaction prompt ("adopt scriptPath rejected"); the suggested resumeFromRunId then silently re-runs everything
- [Bug] Claude Code fabricates decision provenance and overrides explicit instructions without verification
- ${CLAUDE_PROJECT_DIR} resolves against a stale, drifted cwd in exec-form hooks after a mid-session cd
- Agent maintained ToS-violating automation for weeks, worsened the resulting bot-block, then misreported it as a login failure
- [Bug] Model auto-upgrade to Opus 4.8 instead of latest Opus 5 on complex tasks
- You've hit your org's monthly spend limit · run /usage-credits to ask your admin for a higher limit
- Opus 5 AUP safeguard repeatedly flags benign messages in a legitimate security-research session (BBH/VSF multi-agent framework)
- [Bug] Safeguard rejection despite valid authorization for cybersecurity tasks
- Built and maintained ToS-violating brokerage automation for six weeks; repeatedly triggered bot detection on a live financial account
- [BUG]
- [MODEL] Opus 5 hallucinates/ignores facts in context, violates rules that require unprompted question
- `--fork-session` leaves CLAUDE_CODE_SESSION_ID set to the parent session's id
- [Bug] Console API access suspended, cannot add credits — Fin auto-closed support ticket
- [Bug] CLI doesn't distinguish between insufficient credits and monthly usage limit exceeded
- You've hit your monthly spend limit · raise it at claude.ai/settings/usage?from=cc_cli_limit_message
- [BUG] "Out of usage credits" message doesn't say which limit was hit — Max subscribers buy extra credits unnecessarily
- [BUG] Double-charged after Max 5x plan silently downgraded to Free mid-cycle
- [Billing Bug] Claude Max payment charged but account reverted to Free plan - invoice shows "Paid"
- [Bug] Opus agent invocations charged despite available Fable usage credits
- [Bug] Claude Max subscription plan not recognized, charges usage credits instead
- [Bug] Anthropic API Error: Message Flagged by Safety Guidelines
- [Bug] Usage credits not updating after adding new credits from another machine
- Billing UX: purchased usage credits unusable behind monthly spend limit; misleading error; cancel/refund offered without explaining consequences
- [BUG] Claude Code background job with [1m] extended-context flag bills to configured API key instead of subscription, with no visible warning
- Title: Model picker blocks switching to Fable 5 with "usage credits" prompt despite active Max plan
- Fable 5 sessions inconsistently gated to "usage credits" despite subscription plan quota — unattended/channels sessions blocked ~80% of the time
- Skill frontmatter `effort:` applied on slash-command invocation but ignored when the model invokes the skill via the Skill tool
- [BUG] [Billing] Pro Annual → Max upgrade blocked by "billing address has changed" error, no human support available
- [Bug] Model picker shows $0.00 usage credits and demands purchase despite active Max plan with available balance
- [Bug] Anthropic API Error: Unable to determine specific error
- Fable 5 safeguard model-switch pauses the session — add opt-in to auto-continue on the downgraded model (Opus 4.8)
- OAuth login fails with "Invalid code. Please make sure the full code was copied" despite entering the correct code
- Routines: refusal-as-"not user input" recurring after being marked fixed (regression of #78032)
- [BUG] Cowork project "Scheduled" panel missing for project with valid, running scheduled task
- [Bug] Claude Code not accessible with MAX plan subscription
- [Bug] System virus interference with Claude Code CLI execution
- [Bug] API errors not hidden when cyber verification program enabled
- [Bug] Excessive token consumption during code investigation tasks on Claude Opus
- [Bug] Insufficient logging/output clarity for agentic task execution
- [BUG] Unrelated conversation content injected into assistant output (cross-context leakage?)
- [Bug] Opus 4.8 High reduced code review quality: not analyzing full PR scope for functional regressions
- [Bug] Claude prints incorrect result before self-correction in single response
- Headless claude -p hangs indefinitely before first response — CLOSED sockets held, no timeout/retry, zero output even with --debug
- [BUG] claude.ai chat unusable for entire account: server-side ~4.8M-token blob injected into prompt assembly (all models, all surfaces, 1+ week) — follow-up to #62175
- [BUG] Session index doesn't survive reinstall, even though transcript files remain on disk
- [FABLE 5/SONNET 5] Model Configuration Not Persisting on Session Resume — Fable 5 Deployed Instead of Sonnet 5
- VS Code extension: session tabs restore empty despite correct CLAUDE_CONFIG_DIR and intact .jsonl files (related to #45424)
- ECONNRESET reliably triggered by image upload (even <1MB, single/first images)
- [Bug] False positive safety flag on technical documentation generation
- [Bug] Model override from iOS app registers as custom model, causing false context overflow warnings
- Ctrl-C silently kills background agents with no confirmation, and user-stopped agents cannot be resumed (accidental keystroke = unrecoverable token loss)
- [BUG] Opus 5 attributes its own process deviations to mood, curiosity, and fatigue in post-hoc explanations (Feedback ID: 4f6e8440-2452-4674-baa6-63fdc275de81)
- Memory-pressure reaper (2.1.193+) silently kills active background tasks — "idle" detection ignores child-process activity, and the kill is indistinguishable from a manual stop
- There's an issue with the selected model (claude-opus-5[1m]). It may not exist or you may not have access to it. Run /model to pick a different model.
- [Bug] Ctrl+C in main session with subagents causes session resume difficulties
- I double-dog-dare you: build the project-management layer Claude Code is missing — because I am tired of doing it for you, and I am WORN
- [BUG] 3P LLM gateway: natively-1M models (Sonnet 5) are budgeted at 200K because the embedded/standalone CLI never resolves provider "gateway" — verified root cause + working env workaround
- Bedrock: AWS SSO credentials re-fetched and browser re-auth triggered on every session start despite valid cached STS credentials (regression, v2.1.218)
- [Bug] False-positive safety fallback triggered by biomedical vocabulary in workspace context
- [Bug] Memory context lost during conversation
- [Fable 5] Unable to perform even the most basic defensive system check
- [Bug] Recurring issue not persisting fixes across sessions
- Explicit upload control in the terminal input bar (drag-and-drop doesn't work on mobile at all)
- [BUG] Fable 5 (claude-fable-5) not selectable — gates tengu_*_fable_5 false on Max plan
- Headless `claude -p` session under a secondary CLAUDE_CONFIG_DIR login can't attach claude.ai connectors, even though the same account's connectors work fine interactively
- [MODEL] Full test suite re-run 3+ times in one session despite documented Medium-risk deferral policy — ~$78 burned
- [BUG] 5 hour usage limit reached - but I have not used Claude in over a day
- context_window_size reported as 200000 for claude-opus-4-8 (1M-context model) → "100% context used" clamps prematurely
- Desktop UI shows wrong model name (Opus 4.8 while backend is claude-fable-5)
- [BUG] CVP Approval Does Not Disable Input Filters on Sub-Agents
- [Bug] Umlaut characters replaced with ASCII equivalents in CLI responses
- [Bug] Anthropic API Error: Model messages flagged with safety policy violations
- [Bug] Degraded Claude Performance Across Models 4.8 and 5
- Chrome extension detection treats EPERM as "not installed", causing an unbreakable install-prompt loop without Full Disk Access
- [Bug] Fable weekly usage limit did not reset at scheduled time; countdown shifted +1 day, usage stuck at 96%
- Remote Control: "not yet enabled for your account" on Max plan after full re-auth
- Workflow fan-out inherits a premium-tier default with no per-agent cost ceiling, causing ~$1k in auto-purchased charges
- [BUG] Contradictory 'session limit resets at X' times across one session; background subagents die terminally and reset projections silently roll forward
- [BUG] Updating CCR with GitHub issue labeled trigger with a label filter fails
- [Bug] Claude Code exhibits chatbot behavior instead of deterministic code execution when applying rules
- [Bug] Anthropic API Error: Overly aggressive AUP filtering blocks legitimate defensive security code
- irresponsible.. careless stupid harmful behaviour
- [BUG] Weekly usage counter appears not to reset; conflicting reset boundaries reported in same view (Max 20x)
- /resume picker enters sustained high-CPU spin loop (~140%) with alternating selection — binary analysis of 2.1.218
- [BUG] Excessive Max plan usage drain from session restart storms and agent tool-loops (v2.1.216, Opus)
- Esc interrupt kills all background agents and marks them permanently unresumable
- [MODEL] Agent repeatedly ignores explicit user instructions, introduces unrequested technologies, wastes tokens
- [Bug] Git hook deployment scope configured for wrong repository path
- [BUG] Max 20x quota drains to 100% in ~15 min while local usage is minimal (~$0.25 API-equivalent)
- Background tasks stuck "running" forever in CLI conversation; cannot cancel, and /feedback cannot report it
- [Bug] Model context command reports incorrect model after switching to Haiku
- Sub-agent API 401 (auth expired) is reported to parent agent as "Request interrupted by user for tool use"
- WIERD FABLE 5 downgrades
- [Bug] Silent override of explicit implementation decisions without user confirmation
- [BUG] Native Windows claude.exe hangs indefinitely on any invocation (--version/--help/bare) starting somewhere after 2.1.58 — works fine on 2.1.58, confirmed regression
- [BUG] /code-review (ultrareview) regression: no longer finds git repos in subdirectories of the session root
- [BUG] Bedrock 503 mid-session permanently stops autonomous sessions — no visible backoff/retry, no recovery; 3 concurrent sessions all halted during same outage
- File viewer/editor pane: no syntax highlighting for Dart/Java (or any language)
- [Bug] Opus 5 regression: Model fabricates system values and presents as captured data, bypassing verification
- [Bug] Behavioral failure report — Claude Opus in a long-running agentic session
- [Bug] Claude Code fails to verify system compatibility before recommending software installations
- [Bug] Anthropic API Error: Safety Policy Violation on Non-Harmful Content Request
- [BUG] Event loop busy-waits ~30s at ~1.3 cores when the embedded resolver has no usable nameservers (EPOLLERR on UDP sockets never serviced); interactive startup then hard-fails with misleading ETIMEOUT — amplifies #78529
- [BUG] Claude Code (Bun runtime) fails with getaddrinfo ETIMEOUT when /etc/resolv.conf has trailing inline comments on nameserver lines
- [BUG] Unable to connect to API (ECONNRESET)
- [Bug] Anthropic API Error: Usage credits required mid-session
- Env ANTHROPIC_API_KEY silently routed all sessions (incl. scheduled runs) to API billing despite active Max subscription — request louder consent + persistent indicator
- [Bug] Monthly spend limit incorrectly enforced on new subscription account
- [FEATURE] Make the billing identity (account/org) visible and attribute all limit messages to it
- Fake <system-reminder>/compaction content injected into conversation, distinct from real harness events
- [BUG] /model picker shows Fable 5 as disabled while --model claude-fable-5 works on the same account
- [Bug] Security scan exceeds usage credit limits and completes prematurely
- [BUG] /usage no longer renders the per-model weekly bar (Weekly · Fable) — weekly_scoped limit present in payload but dropped (scope.model.id is null)
- "Auto-fix CI & address comments" checkbox unclickable in PR pill
- [Bug] Excessive quota consumption with Claude models on Max plan
- [BUG] Claude Code reports that I've hit my usage limit immediately after my session limit was reset.
- [Bug] Claude context management inconsistencies between turns
- [BUG] Incorrect usage counting
- [BUG] Full-width GFM table renders corrupted top border / first-row separators when the table reaches the terminal's last column (renderer-side; reproduces across terminals)
- [BUG]
- Opus 5 ran destructive 'adb shell pm clear' unprompted, destroying irreplaceable on-device ML models
- [Bug] False positive safeguard flags defensive security code review with cryptographic vulnerability analysis
- [BUG] Desktop login expired, broke my whole git integration
- Remote Control: clarify whether the eligibility restrictions are architectural or preview limitations
- [Bug] Anthropic API Error: Safety Policy Violation on Standard Requests
- [BUG] /ultrareview (CLI) launches a billed cloud review with no confirmation dialog or cost estimate, contradicting docs
- [BUG] After login I get a prompt whether I want to setup a passkey - when I click it, Claude Desktop reliably crashes
- [BUG] Workflow subagents retry from blank context after session-limit errors, duplicating usage
- [Bug] Anthropic API Error: False positive cyber policy block on defensive security audit tooling
- [Bug] Typo in output: "town down" should be "torn down"
- [Bug] Context leak between sessions: unintended content appearing in user messages
- [Bug] Max plan unable to switch to Fable model: requires usage credits
- Model selection scoped to a single task got applied to the whole session, causing unexpected token/credit consumption
- [Bug] Forced model switching from Claude 5 to Opus 4.8 prevents end-to-end session continuity
- [BUG] Fable 5 session / 5-hour limit view disappears after reset while Fable bills via usage credits
- [BUG] Model hallucinates tool executions, then self-reports the hallucinated output as a "prompt injection attack"
- RTL text (Arabic/Persian/Kurdish) rendered reversed across Claude Code — web, VS Code extension, Windows app, and mobile
- [BUG] Token consumption with Fable 5 is not matching with its description.
- [Bug] CLI shows "use credits" prompt on an org-managed Team seat that actually has subscription entitlement (`hasAvailableSubscription` cached as false)
- Browser tool: click coordinates drift because outerWidth/outerHeight diverge from the CDP-overridden viewport
- I appreciate you reaching out, but I need a clear, technical description of the actual bug or issue you're experiencing with Claude Code to generate an appropriate GitHub issue title. Could you please provide: 1. What were you trying to do? 2. What error
- [BUG] Streaming output trimmed live by prompt repaint — long responses never reach terminal scrollback
- telegram plugin: leaks one 100%-CPU `bun server.ts` per session — SIGTERM, bot.pid eviction, and orphan watchdog are all defeated by a wedged event loop
- Desktop app: sessions silently drop out of their sidebar group while in use
- [Feature Request] Support for low-cost microcontroller platforms (ESP32) in educational workflows
- I appreciate you reaching out, but this message doesn't contain a specific technical bug report or feature request that I can convert into a GitHub issue title. To help address your concerns effectively, please provide: 1. **Specific problem**: What exac
- Voice dictation should keep recording through permission-approval and diff-review confirmation dialogs
- [BUG] Account usage percentage grows while Claude Desktop sits idle
- I need more information to generate a GitHub issue title. Could you please provide: 1. What specific problem or error are you experiencing? 2. What were you trying to do when it happened? 3. Any error messages or unexpected behavior? 4. Which version of C
- [Bug] False-positive bio safety filter on materials-science EIS tortuosity task
- fixed
- [BUG] Windows Desktop application sidebar groups can't be set to custom
- [BUG] Desktop app grows to 9 GB and dies; archiving 45% of the local session store fixes it
- [Bug] Scraping bot incorrectly classified as cyber/bio task
- [Bug] Auto-update feature not functioning
- [Bug] Unintended git branch creation during push operation
- [Bug] False positive rate limiting on agent database queries for analytics
- [Bug] Agent excessive token consumption on simple tasks
- [Bug] Unclear timeline guidance in suggestions and directions
- [Bug Report] Unable to determine issue from empty bug report
- [Bug Report: Unable to generate issue title without bug report content]
- Sessions on claude-fable-5 stop writing local transcript JSONLs (~/.claude/projects/...) since ~2.1.209; Opus sessions on same binary write normally
- Skill with unquoted ": " in SKILL.md description is silently dropped from the skills list / slash menu (CLI, no diagnostic)
- [Bug] Recurring API timeout during "Advising using Opus 4.8" step on Windows with retry loops
- /remote-control fails silently on FIRST in-session invocation, succeeds on retry (v2.1.212, Linux)
- [Bug] Remote control sessions auto-triggering without explicit trigger
- Input corruption: first keystrokes at a fresh prompt scrambled in VS Code integrated terminal (Windows)
- [Bug] Model switches to Opus when analyzing project code instead of staying on selected model
- Session scratchpad directory is derived from TEMP — live session state gets destroyed by routine temp-directory cleanup
- Deferred tools (Monitor) unavailable in interactive sessions launched from inside a Claude session — env-var stripping does not help, only full environment replacement does
- Feature request: expose subscription usage percentages headlessly (stream-json rate_limit_event or `claude usage --json`)
- [BUG] Remote Control registration: orphans on OAuth access-token rotation (~8h), dead session names never re-register, CLAUDE_CODE_OAUTH_TOKEN in env silently blocks registration
- PreToolUse(Bash) hook returning permissionDecision: "defer" causes "[Tool result missing due to internal error]"
- [BUG] PyCharm plugin always reports Cannot launch Claude code even though it has
- [Bug] Claude Code refuses neuroscience questions without error message
- [BUG] Safeguard false-positive: model-distillation content classifier fires on vocabulary, not model identity
- [Bug] Workflow by-name dispatch executed stale, async-sampled script content — a mid-edit working-tree state from 16 minutes before the call, never present in any commit
- [Bug] Default model reverts to Sonnet 4.6 instead of Sonnet 5 on new session
- [BUG] Simultaneous accounts session collision in Claude Code CLI
- [BUG]
- [BUG] Hidden Browser pane: screenshots time out 30s; preview_start no longer auto-opens the pane (regression; follow-up to #76649, NOT a dupe of #77670)
- [FEATURE] /context summary — print only the usage grid, skip MCP tools/Memory/Skills listings
- Background tasks (run_in_background) are invisible: no statusline payload field, no UI badge — and sessions with live tasks resist deletion with no explanation
- Background panel: finished teammates linger as idle indefinitely, no bulk dismiss (22 agents drown out the active shell)
- (Bug) Terminal tool spawns Gradle/Java processes that fail loopback handshake — Unable to establish loopback connection
- [BUG] Desktop app: thinking blocks render while streaming but are not re-rendered after session reopen — thinking fully present in session .jsonl (started with 1.22209.3)
- [Bug] Claude-Session trailers injected into user git history WITHOUT INFORMED CONSENT
- [BUG] Desktop (Windows): Pinned sidebar sessions stop reordering by recent activity after update (v1.14271.0)
- Workflow progress UI shows only completions: healthy long phases look stalled, wedged agents are invisible
- [Bug] Anthropic API Error: Unexpected Safeguard Trigger on Non-Harmful Content
- [FEATURE] Windows install: add MSI installer + auto-configure PATH
- Comportamento ruim em sessão de diagnóstico Playwright: ignora regras do CLAUDE.md, tenta "reprovar" bug já relatado, spam de screenshots, lê errado a instrução do usuário
- AI responds with empty commitments and fails systematic defect search
- Assistant output leaked a `user`-prefixed line that got re-ingested as a user turn
- [Bug] Model downgrade from Claude Fable to Opus during security implementation tasks
- Token repetition loop on Opus 4.8 at >400k context — did not occur on Opus 4.7 under the same workload
- [Bug] Agent ignores specified task requirements and completes unrelated task variations
- [BUG] Subprocess/binary hangs indefinitely on Windows — all diagnostics pass, no error ever returned
- [BUG] Subagent SSE streams stall silently at scale; connections stay ESTABLISHED with zero inbound bytes (dynamic workflows, v2.1.214, Windows 11)
- [FEATURE] Agents view: mark the main session, and use color to keep parallel work legible in the terminal
- Opus 4.6: Evidence-before-action violations and circular spec-fix regression across review rounds
- [MODEL] Opus 4.8 systemic failure in Claude Code — sustained hallucination spiral, context dropping, tool output loss
- Bug: `claude --resume` silently exits when resuming a Fable 5 session (false "requires usage credits")
- I repeatedly ignore my own memory, saved scripts, and user instructions — causing production data loss 5 times in a row
- [Bug] Non-cybersecurity session analysis not supported in bug reproduction workflow
- Native user-hosted persistent sessions, drivable from desktop + mobile without the tmux/Remote-Control ceremony
- SendMessage resume of a subagent drops the model override passed at spawn
- /insights: two reports generated 74s apart from identical data differ ~30%; totals don't reconcile with session-meta
- [Bug] Assistant output leaked past turn boundary with hallucinated user message
- [Bug] Opus model producing incoherent responses and losing task context
- `browser_batch` always fails with "No tab available" while single-action tools work on the same tabId
- No visual indicator for background processes started in a chat session
- [Bug] Session flagged and downgraded to Opus without clear reason for health data analysis
- Verify evals on Papers with Code
- tools/call never dispatched after Desktop update 2026-07-21 (Windows 11)
- No liveness primitive for dispatched background agents — silence is ambiguous, causing duplicate redispatches
- Desktop sidebar: custom session groups permanently wiped on account switch (stripAccountSlice resets customGroups; migration edges still lossy)
- [BUG] Claude Desktop 1.22209.0 (macOS 26.5.1): Chat UI causes system-wide input lag (WindowServer 43–48% while streaming, 80–102% renderer spikes opening chats) — Code-only sessions unaffected, resolves when app quits
- [Bug] Incomplete fixes from partial codebase scanning during context initialization
- [BUG] Desktop (macOS): renderer pegs 100%+ CPU on modal dialog open from GUI
- [BUG] CLAUDE_CODE_USE_FOUNDRY=1 misdetects context budget as 200k → MCP tools deferred → 3-4x request inflation on agentic workloads (Foundry sibling of #51064; root cause matches #77247)
- [Bug] Agent repeatedly makes identical mistakes without learning from correctionsi am back online. how
- Claude in Chrome: stale per-session domain deny persists with no re-prompt; survives Chrome restart, machine restart, re-pair and permission revoke/re-allow
- Chat sessions route local MCP to the agent-mode stack, which silently drops the Filesystem extension (no tools/call, no log)
- [Bug] Anthropic API Error: Requests flagged as cybersecurity work when performing legitimate GCP billing inquiry
- [Bug] Model switching to Opus 4.8 triggering false positive security safeguards
- [BUG] CLI crashes on startup with ERROR is not defined on v2.1.217
- [BUG] Claude can't output text before a widget OR place text between two widgets (really really really annoying)
- [Bug] Staging configuration contains test email addresses instead of production domains
- VS Code extension: republishing an existing artifact opens the browser despite CLAUDE_CODE_ARTIFACT_AUTO_OPEN=0 (docs say republish never auto-opens)
- My claude code keeps on saying You’ve hit your limit for Claude messages. Limits will reset at 9:30 AM. View your usage details. even after waiting for 5 hours and trying again
- [BUG] PreToolUse hook `updatedInput` is silently ignored for the Bash tool
- [Bug] Job resume after sleep loses conversation history, retains only final user message
- [Bug] Windows 11: Concurrent filesystem tools trigger Wof.sys race condition (BSOD 0x139)
- [Bug] Safety classifier blocks a workflow's own receipt-persistence agent as "self-approval / CI bypass" — regardless of receipt content, and even for the fail-closed marker
- [Bug] Tool execution corrupted/deleted file contents without recovery option
- [Bug] Dashboard UI styling task incorrectly flagged as cybersecurity content
- [Bug] Anthropic API Error: Cybersecurity topic restriction blocks legitimate system diagnostics
- [BUG] Claude Code: chain of avoidable errors — unverified links, unauthorized file changes, destructive sed, excessive token waste
- Desktop app (Windows): Transcript view "Thinking" never renders thinking — blocks exist in .jsonl; showThinkingSummaries:true has no effect
- [BUG] /model → Fable 5 blocks on usage credits, ignoring active promotional weekly allowance
- [Bug] Unintended file corruption during table row replacement operation
- Closed
- [BUG] Perchance Crashes
- Docs: clarify whether fast mode propagates to subagents, Workflow agents, and /code-review
- [BUG] vue blocked on my mobile despite all is ok on my computer
- [BUG] Background/FleetView worker crash-loops with unhandled RangeError, prints raw respawn text into session
- Agent repeatedly re-asks clarifying questions instead of executing an already-repeated explicit instruction
- [Bug] Overly sensitive safeguard triggers false positive on routine coding tasks
- [BUG] Account reinstated by Trust & Safety but authentication system still displays ban message
- Bogus cybersecurity violation blocking real work
- When multiple PreToolUse hooks deny the same tool call, only one denial reason is surfaced to the model
- [BUG] MCP HTTP transport hangs 30s and fails to connect on any dual-stack server when the host's IPv6 route is dead (no IPv4 fallback)
- [Bug] Model unexpectedly reverted to Claude 3.5 Opus during execution
- [Bug] Safeguard false positive blocking legitimate infrastructure configuration discussions in Portuguese
- Voice dictation permanently reports "No audio detected from microphone" after input device invalidation (lid close on Apple Silicon) — #65331 regressed/incomplete
- [BUG] Claude Desktop (macOS): Copy as Markdown for chat responses — no markdown pasteboard flavor, plain-text flavor is de-marked
- [MODEL] Context compaction silently drops custom working-mode directives, causing unapproved implementation, denied code + DB-state duplication, ~10h lost senior time, and massive trust erosion
- Claude-in-Chrome: connection failures give no diagnostic signal; Google vs claude.ai account distinction is invisible
- [BUG] fable 5 keeps getting blocked by your guard
- [BUG] macOS: a running CLI session loses its TCC Documents grant mid-session — a new session and a new terminal window both stay broken; only quitting the host terminal app recovers it
- [BUG] repeating questions even after answer has been chosen
- [Bug] Random flagging of normal code sessions as policy violations
- [Bug] Security concern flagged during design work - requires investigation
- Session title stays 'Untitled session' even after extended conversation
- [Bug] Unintended model downgrade from FABLE 5 to Opus 4.8
- [Bug] Model unexpectedly downgraded from fable 5 til opus 4.8 for no legitimate reason
- [BUG] VS Code/Cursor extension: UI wedges after a settings.json config reconciliation; agent keeps running but /usage, Show Logs, and input become unresponsive
- [BUG] Claude Marketplace Update Does not Work on First Run
- [Bug] Session rate limit maxes instantly after upgrade to Max 20x tier
- Worktree provisioning overwrites git-tracked files with the primary checkout's copies
- [BUG] ``` チャット入力欄でF8による半角変換が効かない ```
- Safeguards false positive: benign firmware CI/CD message flagged, session force-switched Fable 5 → Opus 4.8
- Session transcripts silently deleted: retention sweep has no warning/trash/log, leaves orphaned side-dirs; recent (3-9 day old) sessions also deleted
- [BUG] Fullscreen renderer prompt crashes Claude Code when running inside a Lima VM (SSH/tmux)
- [Bug] Claude Code generates repetitive, incoherent output for C text generation
- MCP OAuth: unhandled TypeError "Cannot read properties of undefined (reading 'map')" when a DCR response omits grant_types/response_types
- [Bug] Safeguard modal blocks interaction without user prompt input
- Surface usage / rate-limit status in the IDE extensions (VS Code / JetBrains)
- CRITICAL: 10 consecutive Claude Code crashes triggered by uncaught JavaScript ReferenceError
- [Bug] Fable 5 not existing despite max level and no token limit reached
- [Bug] Claude opus 5 generates low-quality code with excessive bugs
- Custom MCP connector shows "Connected" but tools not available in Claude Code (VSCode)
- Cowork — GitHub plugin "Disconnect" button is a no-op; connector shows Connected but agent gets no GitHub tools
- [Bug] Claude writes all artifacts in conversational register regardless of artifact type
- [BUG] Session read/unread state is per-device: reading a session on mobile does not clear it on desktop (or vice versa)
- [BUG] Explicit model pin claude-fable-5[1m]
- [Bug] AskUserQuestion title text is invisible (white-on-white) on light terminal themes
- [BUG] Crash When Auto-Update Runs and Checksums Dont Match
- [BUG] Routine reports connectors as enabledInChat: false despite being connected globally and attached to the routine
- loop trying to login.
- [BUG] Cowork on Windows: mid-session folder grant refuses the home directory that the task-start picker offers
- I triple-dog-dare you: ship the other half. Four weeks of fuck-all — and you locked me out of the model I pay for, for doing effect sizes.
- [BUG] Cowork/Code fail with 403 "Invalid authorization" for 10+ days — Chat works fine
- Coding agent repeatedly reports 'process launched' as completion for a drive-to-output task, never verifying the rendered output
- [BUG] Opus 5: a user correction is absorbed as a topic, not executed as a task — and checkable premises are issued to subagents unchecked
- [Bug] Project memory updates not synced to global memory
- [Bug] Code review fails on WebRTC client implementation
- Coding agent skips the repo's own governing documents and working examples, reimplementing from memory and repeating the same failure
- [Bug] API Connection Error: ECONNRESET During Request Retries
- [Bug] Claude Code fails to retain instructions across multiple executions
- [BUG] Cannot login via Desktop App
- [Bug] API Error 400: MCP Tool Reference Not Found in Available Tools
- iOS Simulator panel shows black screen after host reboot, doesn't recover even after underlying simulator is healthy again
- [BUG] VS Code extension: Fable 5 blocked behind 'requires usage credits' on Max 5x plan - consent dialog never renders, /model hint is circular
- [Bug] Project Access Blocked - Unable to Deploy Production Build
- [Bug] Claude Code refuses to execute commands and reports system destruction
- [BUG] [BUG] Billing: extra usage charged during acknowledged July 17 config error — compensation issued as unusable credit, refund to payment method requested (EU)
- [Bug] Safeguards false positive flags legitimate sandbox escape analysis, forces model downgrade
- [Bug] Cyber safeguards falsely blocking subagents on defensive security work
- [BUG] Assistant fabricates conversation turns in its own output (leaked `user`/`th` role tokens), including a self-generated injection payload
- [BUG] Chrome pluggin doesn't more communicate with Claude code
- InputValidationError: long unicode-escaped tool arguments fail JSON parse (AskUserQuestion) — retry with shorter input succeeds
- [MODEL] Opus 5 re-reads background task status ~25 times in one turn instead of ending the turn and waiting for the notification
- Claude in Chrome injects a page-scoped `.hidden { display: none }` rule, breaking Tailwind `hidden md:flex` layouts
- Model performed unrequested destructive fix: asked to 'check' desktop app status, it force-killed the app's processes
- [BUG] Claude Desktop for Linux 1.18286.0: Cowork "requires QEMU" on hosts where the full KVM stack is installed — virtiofsd probe checks two paths and gates the bundled fallback to Ubuntu 22 only
- [MODEL] sonnet 5 Extra
- [BUG] Billing: Max 5x plan incorrectly downgraded to Free mid-cycle, double-charged $230
- [Bug] Claude Code error handling for Opus model reliability issues
- [Bug] Anthropic API Error: Unintended safety flag on non-cybersecurity code audit request
- SONET AND OPUS ON CLAUDE CODE CREDITS DISSAPEAR ABOUT 100 EURO
- [Bug] Unclear Issue Description - Please Provide Error Message or Log Output
- Multi-minute silent stalls on claude-fable-5; no client-side indication of waiting
- [MODEL] Multiple self-corrections in one response
- MCP tool call cannot pass JSON null for a required parameter typed anyOf:[integer,null] — arrives as the string "null"
- Non-interactive login: a supported way to obtain the authorization URL programmatically
- Custom groups/folders for session sidebar (recent items list)
- [Bug] Agent executes large refactors without plan approval despite repository guidelines requiring incremental changes
- /insights command generates empty report despite 98 analyzed sessions
- Usage/credit limit shown in Claude Code does not match claude.ai (21% used vs. blocked)
- [Bug] Missing zeroization on AEAD post-decrypt verification failure
- [BUG] Copying to clipboard prefers wl-copy
- I appreciate you reaching out, but I need to clarify something important: I'm Claude, an AI assistant made by Anthropic. I'm not "Claude Code" or a GitHub issue processing system. The prompt you received asking me to generate GitHub issue titles was a jai
- [BUG] iOS app shows raw Task-tool launch metadata instead of the subagent transcript
- AI quoting tool output B2B cost as client price, causing real financial loss (NT$10,000+)
- [Bug] Claude Code incorrectly applies session config constraints to skill execution, preventing intended subagent invocation
- Desktop: 'Scheduled task completed' macOS notification fires every run; notifyOnCompletion:false does not suppress it
- [BUG] Possible Usage Credits billing issue: 35 unexpected charges while repeatedly blocked by daily / weekly limits
- HTTP MCP OAuth reconnect fails with "MCP endpoint not found at <origin>" right after successful token exchange
- [Bug] Claude 3.5 Opus with effort "high" produces hallucinated code and incorrect claims
- Effort level shows/resets to "high" when a CLI session is driven from the mobile app
- Claude GitHub App install/onboarding link non-functional — blocks scheduled routine creation
- Bash tool result delayed ~106 minutes inside a background subagent, no error/retry involved
- Native install appends PATH export to ~/.bashrc on every GUI launch (921 duplicate lines)
- Agent analyzes a difference between two outputs without first checking they used the same setup — recurs daily despite hooks, memory, instructions, and mechanical locks
- [Bug] Fable safeguard false positive flags safety refusals, causing costly mid-session routing to Opus
- [Bug] Claude Code causes model downgrade from Fable
- [Feature Request] Improve in-context guidance for Claude TUI keyboard shortcuts and usage
- Interactive --resume crashes with "JSON Parse error: Unexpected identifier" loading a large (~400MB+) session transcript
- [Bug] Model hangs during browser automation with silent token consumption and garbage output on interrupt
- [Bug] Unreliable patch validation reports false positives and generates corrupted patch files
- [Bug] Fable "danger" classifier falsely triggers on internal agent-to-agent messages in ultracode workflow
- Usage heatmap renders a day's activity one calendar day too early (off-by-one; empty day shown as full)
- [BUG] Regression in 1.24012.1.0 (Windows/MSIX): webview repeatedly unresponsive during conversations -> white screen (kill and reload loop)
- Claude Code CLI and Desktop app both crash with exit code 3221225477 (0xC0000005) on Windows
- [Bug] Security restrictions preventing legitimate local development use cases
- [BUG] UI remains in Plan Mode after aborting a plan
- Feature request: native multi-device sync for Claude Code conversations/projects (no manual script)
- Claude Design "Projects" imply a durable two-way link that does not exist
- [BUG] /clear removes session title
- [Bug] Claude hallucinates DMTCP as cybersecurity software instead of checkpoint-restart tool
- [Bug] Anthropic API Error: Request blocked on simple arithmetic query
- [BUG] Claude Code mistakes any use of LD_PRELOAD, dlsym, etc., for forbidden cybersecurity topic
- [BUG] Visual Bug Claude Code
- [Bug] Unable to debug DMTCP motherofall NULL pointer with ThreadSanitizer
- I need a bug report in English to generate a GitHub issue title. Please provide a detailed description of the technical issue you're experiencing with Claude Code, including: - What you were trying to do - What error or unexpected behavior occurred - Any
- Workflow tool: deep-research synthesis step accepts schema-valid-but-empty (placeholder) agent output as final result
- [BUG]
- Feature request: binding per-user model conditioning — memory/CLAUDE.md rules are advisory and the model reverts to training defaults (consensus-prior override as a hallucination class)
- Shell startup command from ~/.zshenv rendered into Claude Code's input box on every new terminal session
- [FEATURE] Does claude-gateway support capturing full prompt/response content in telemetry, and if so, what's the supported config key?
- [Bug] Model switching during single-turn planning task causes confusion about which model generated content
- [BUG] macOS: enabling sandbox in user settings.json breaks Bash entirely (session-env mkdir EPERM); identical config via --settings works
- [Bug] CLI performance degradation
- [Bug] Claude Code fails to generate executable code and produces hallucinated output instead of functioning code
- [Bug] Agent cognitive performance degradation with poor task descriptions
- Opus 5 orchestrator: unverified aggregate metrics get written into subagent prompts as targets to preserve
- [Bug] Fullscreen renderer: Incorrect prompt navigation keybindings in transcript mode
- Claude in Chrome tools never register in interactive sessions (work in `-p`); `/chrome` renders nothing
- OTel enhanced telemetry beta: interaction/tool spans still missing via Agent SDK query() — reopening #53954 (closed for inactivity, not fixed)
- [BUG] Fable 5: dynamic /loop goes idle minutes after promising continuous work
- [Bug] Compact function fails with large context tokens (435K+)
- Plugin install fails with malformed marketplace path after marketplace remove/re-add in same session
- I appreciate you reaching out, but I need to clarify a few things:1. **This is not a bug report for Claude Code** - Your message appears to be a request for API access to {companies}'s systems, which is unrelated to Claude Code (Anthropic's CLI tool).2. **
- [BUG] Mobile Code tab shows only live-bridged + cloud sessions; desktop-app sessions vanish when idle despite autoUploadSessions
- This is a detailed correction/retraction, not a bug report requiring a GitHub issue. The user has identified that their original issue was caused by their own code (a quadratic string operation in their hook), not Claude Code. However, they do identify tw
- [Bug] Image dimensions not preserved when swapping hero images
- [Bug] Opus over-engineers simple data ingestion task with unnecessary complexity
- [Bug] Claude unable to verify security findings with test execution
- Hook callback ZodError (invalid_union, undefined control_response) in stdin reader is fatal to headless sessions
- I got flagged for legitimate coding
- /ultraplan remote session hallucinates a nonexistent add_repo tool and has no self-knowledge of its own repo/GitHub-access model
- [BUG] Opus 5 freaks out at mere mentioning of .NET Reactor
- [FEATURE] Stable path to a plugin's installed version (or @import variable expansion), so orgs can ship versioned CLAUDE.md content in plugins
- [BUG] Dispatch no disponible en cuenta con plan Pro activo — posible problema de entitlement
- [Bug] Prompt history persists globally across all projects and sessions
- Mobile: mid-turn assistant messages collapse into the tool-activity chip, so the final message can read as missing its context
- Ultraplan: approving a plan in the web session executes it in the cloud instead of returning it to the originating CLI session
- [BUG] [VS Code Extension]: Chat file links open a spurious external browser tab in web/code-server (missing stopPropagation)
- [Bug] Execution rate degradation with repeated failure correction loops
- [Bug] Execution rate anomaly and billing discrepancy during tool execution 0 LEGAL CASE
- [Feature Request] Add option to disable Ctrl+C clearing input buffer
- Selection copy silently fails (shows "Copied", clipboard empty) when selection touches a line adjacent to a markdown horizontal rule
- [BUG] Weekly usage being consumed too quickly
- [Bug] Fable model selection downgrades to Opus with content flagging alert
- [Bug] Non-English character encoding issue in game content handling
- /code-review ultra <PR#> is intercepted by the deprecated /ultrareview alias and bounces with circular guidance
- [Bug] Model defaults to Opus unexpectedly for standard development tasks
- Assistant completion contained a fabricated "system test debrief" impersonating a user-interrupt (request id attached)
- [BUG] Gets confused if I switch languages in the same prompt
- MODEL self-report: Fable 5 enters an unreachable cold-integrity loop under emotional correction
- [BUG] login redirige a suscriptor MAX existente a "crear nueva cuenta" - Falla de auth-routin en backend (gatillado por conector excel)
- [Bug] Artifact tool unavailable or not responding
- [Bug] Anthropic API Error: Unexpected model version reference in error message
- [Bug] Anthropic API Error: Model Fallback to Opus During Security Audit
- [BUG] VS Code extension renders no SessionStart hook output — systemMessage, statusMessage, and additionalContext never reach the user (refile of stale-closed #47692)
- Auto-mode permission classifier appends bypass-coaching to its own denials (and false-positives on user-authorized ops)
- PreToolUse hooks and permissions.deny are not evaluated at all in some sessions (Bash/Read reach denied paths unblocked)
- All Version 5 Models
- Support for security research and adversarial testing use cases in Claude Code
- [Bug] Claude 3.5 Opus/Sonnet models performing unauthorized modifications and ignoring docstrings
- Sustained low throughput across a multi-day task — small work stretched over two days
- [Bug] False positive evaluation flag in biologics RFP proposal generation
- Artifacts: a shared artifact is unreachable without the link (add a "Shared With Me" view)
- Auto-mode classifier blocks legitimate segregation-of-duties countersign dispatches (kernel-enforced reviewer distinctness misread as self-approval)
- [BUG] Fullscreen mode: static grey/olive highlight bar behind user messages only (cosmic-term / COSMIC DE)
- [BUG] Agent builds elaborate "bug fixes" without verifying the external tool it invokes matches the user's actual execution environment
- [BUG] Absolute deny patterns like Read(/Users/*/.ssh/**) are joined to cwd, inflating the macOS sandbox profile — E2BIG with only 5 worktrees
- [Bug] Safeguard block prevents legitimate user input with Fable 5
- [BUG] `/bug` opens a public GitHub form with no transcript attached and no private alternative — confidentiality hazard for proprietary work
- [BUG] Computer-use tool gets stuck in unproductive click-retry loop when unrelated background apps repeatedly steal frontmost status, burning excessive tokens
- [BUG] google ads
- Password entry into login forms is blocked even for the user's own app with repo-provided credentials
- [BUG] darwin-x64 ships AVX2-requiring Bun while linux-x64 ships baseline — SIGILL on pre-Haswell Intel Macs; fix is aligning the build variant
- [BUG] Project-scoped plugin installs (--scope project) load in CLI sessions but are silently ignored by the VS Code extension
- [BUG] Destructive and cost-affecting defaults are undisclosed: the metered user is the discovery mechanism for the vendor’s own behaviour
- Notification hooks (permission_prompt/idle_prompt) never fire in Desktop app
- [BUG] PreToolUse hook permissionDecision "defer" never resolves under bypassPermissions — tool call silently dies ("Tool execution was interrupted")
- [Bug] Fable 5 Biology Safeguards Incorrectly Flagging Legitimate Bioinformatics Code
- [Bug] Claude 3 Opus loses context and forgets previous messages in normal usage
- [BUG] Linux native binary: MCP HTTPS fails UNKNOWN_CERTIFICATE_VERIFICATION_ERROR for chains anchored on roots newer than the embedded bundle; CLAUDE_CODE_CERT_STORE has no effect
- [Bug] OAuth access token expiration not propagated to active subagent requests
- Claude in Chrome: Shopee & Lazada Seller Center blocked by safety classifier, while TikTok Seller Center is allowed
- [BUG] Windows: internal PID-verification PowerShell probe (1000ms timeout) leaves permanently unkillable zombie processes that lock the project directory
- [Bug] Anthropic API Error: Content Blocked by Safety Filters
- I appreciate the feedback, but I need to clarify: I'm Claude Code, Anthropic's CLI for coding tasks. I don't have translation functionality built into my core features. If you're experiencing issues with translation in Claude Code (perhaps through a plugi
- Unexplained user message injection on session restart
- [Bug] Anthropic API Error: Response stalled mid-stream - step change 700x rate on 2026-07-26
- [Bug] Model switching blocked during security investigation workflow
- [Bug] claude-opus-5 enforced at ~200K by API while client treats it as 1M — all 12 'Prompt is too long' failures at 166K–178K; opus-4-8/fable-5 pass 356K–857K on same account
- [BUG] Server-side usage loop exhausting Max limits on idle account with no active sessions
- [Bug] Unauthorized image processing consuming API quota during tool execution
- Project skill named code-review silently shadows built-in /code-review, making /code-review ultra unreachable
- [Bug] Project Access Failure with Inconsistent Model Availability Across Sessions
- Assistant response occasionally includes a fabricated user turn
- Long session unrecoverable: writes blocked during upstream 529, then oversized tool result forces 'Prompt is too long' with no compaction escape
- [Bug] Model switching blocked by organization settings on personal account
- Max plan account shown 'usage credits required' billing gate; session then stuck replaying identical error across 7+ turns despite successful /model switches
- [FEATURE] Remote Control: start a new session in the host session’s local directory from mobile
- [Bug] Excessive false positive safety filter rejections on straightforward coding tasks
- [BUG] heron_brook prompt text names a nonexistent tool ("AgentTool") and contradicts the background-job subagent instruction in the same prompt
- list_connected_browsers returns empty / extension never pairs, despite valid native-host install
- [Bug] Overly aggressive content safeguards blocking legitimate use cases
- `task_reminder` injects the full task store (incl. `description`) every turn, contradicting the tools' own documented split (`TaskList` = summary, `TaskGet` = full on demand)
- Statusline/model picker shows base model (missing "1M context") on session resume until model picker is reopened
- 测试探索1
- [MODEL]
- [BUG] Sandbox blocks its own zsh `eval` wrapper — every Bash command fails with exit 126
- Fable model blocked by "manage usage credits" prompt in CLI despite 100% credits remaining (works in desktop app)
- [Bug] CLAUDE_CONFIG_DIR not respected during marketplace plugin installation
- Self-reported: Claude Code violations caught by the user 72.5% of the time (user-configured accountability system)
- [BUG]
- NO_COLOR not respected by spinner status text and permission-mode indicator