Digest / July 12, 2026
Claude Code Issues Digest — July 12, 2026
223 new issues filed ·
160 resolved in anthropics/claude-code (UTC day).
New issues (223)
- [Suggestion] Expose usage-limit reset time to the agent so it can self-schedule retries
- Background jobs auto-commit/push/open PRs by default, overriding user CLAUDE.md git rules
- Silent model fallback: Fable 5 -> Opus 4.8 mid-session with no notification when a usage limit is hit
- Mobile: always-on context-window usage indicator (could reuse the 'Create PR' action area, no extra real estate)
- [BUG] Bash permission matcher misparses operators inside quoted arguments — a quoted | in a grep pattern defeats Bash(grep:*) and forces a prompt
- 정상적인 코딩 작업 중단
- [Bug] Security triggers still triggering in accepted accounts
- [Bug Report] Model not functioning as expected
- [BUG] Scheduled tasks: clicking "Run now" silently does nothing (no run session, no lastRunAt, no error)
- [Bug] Fable 5 safeguards over-triggering on legitimate personal device configuration, causing unwanted model downgrade
- Add ability to expand and edit collapsed pasted text in the prompt
- [BUG] ConnectionRefused on /v1/messages streaming while simple API requests succeed (Windows, Bun)
- I appreciate you reaching out, but I think there may be a misunderstanding. I'm Claude Code, Anthropic's CLI tool for agentic coding assistance. I'm specifically designed to generate GitHub issue titles for bug reports and feature requests related to Clau
- Windows (no WSL): extension triggers WSL install prompt at every VSCode startup via wsl.exe wslpath call
- サイドバーの「co-chat」「Co-work」タブが表示されない。
- [BUG] Windows: preview_start with {url} returns navOk:true but tab is empty; downstream tools error "No site is open in this tab"
- Esc interrupt kills all background agents and marks them permanently unresumable
- Hooks fail open when CWD is invalidated (e.g. worktree deletion)
- [FEATURE] WebMCP tool discovery in Claude in Chrome — revisiting #30645 with production implementation data
- [BUG] Bug de renderizado al usar scroll táctil en la terminal de Claude Code
- Remote Control: environment lifecycle — stale environments accumulate, are indistinguishable from live ones, and cannot be removed
- Remote Control: SessionStart hook sessionTitle output is delivered to the bridge but never applied to the session's display title
- [BUG] Session-start repo/branch picker ignores the GitHub repository's configured default branch
- Background processes leak as orphans (kill %1 silently fails in non-interactive shell)
- [BUG] spawn ENAMETOOLONG on every local session - Windows 11, MS Store build 1.20186.1
- [BUG] Scroll-wheel SGR mouse report split across stdin reads leaks into prompt input (50ms incomplete-escape flush timer force-emits the partial)
- [Bug] Overly restrictive content filtering blocks legitimate AI research
- skill-creator plugin: trigger-eval harness false-negatives (shared probe-root contamination, first-tool-only detection, 30s timeout)
- [Bug] Subagent returned adversarial security payload instead of code review result
- claude on ARM
- [BUG] iOS Remote Control: project/folder selection step disappeared when creating a new session on a connected Mac — sessions always start at $HOME
- [FEATURE] PreExitPlanMode hook — intercept before ExitPlanMode call for planning artifact adversarial audit invocation
- Tool-result stream injection: fabricated git output attempted to induce destructive git reset
- [BUG] Safety classifier too sensitive - switches to Opus 4.8 for harmless commands like /feedback
- [BUG] Entire ~/.claude deleted while client idle (2.1.207, Windows)
- [BUG] Windows desktop app: 'Usage limit reached' persists ~18h while Usage page shows 0% session / 27% weekly / 51% Fable (Max 20x)
- [BUG] Cowork tab missing on Windows 10 Pro 22H2 — CoworkVMService running, logs folder never created, clean reinstall does not fix
- [Bug] Claude Web's GitHub API proxy also blocks GitHub Copilot
- [BUG] Session transcript never created: new interactive sessions run for hours writing no JSONL and no history.jsonl entries (silent, unrecoverable conversation loss)
- [Bug] MiniDisc playback not supported via PC connection mode
- [Bug] False positive cybersecurity flag on legitimate local development tasks
- [Bug] Content filter blocks legitimate legislative document analysis
- [Feature Request] Reduce Claude 3.5 Fable safety filter false-positive rate for legitimate code reasoning tasks
- [BUG] Claude Code mistakes any use of LD_PRELOAD, dlsym, etc., for forbidden cybersecurity topic
- [Bug] Incorrect model name "fable" - not an Anthropic model/model
- [Bug] Model downgrade to Opus when attempting security fixes in codebase
- Claude Code keeps auto-retrying indefinitely after hitting usage/token limit instead of stopping
- [BUG] TUI renderer + stdin die mid-session while the engine keeps running (macOS, 2.1.207) — idle CPU, no SIGWINCH redraw, transcript still being written
- [Bug] Incorrect Content Policy Flag on Safe Development Code
- Dispatch: interactive option dialogs (AskUserQuestion) don't appear on synced mobile app, blocking remote sessions
- [BUG] Routines: no way to list/reopen a routine's session in the mobile app after the notification is gone
- Allow intentional empty/whitespace-only assistant responses without the "no visible output" nudge
- [BUG] iOS App blocks session initialization with "GitHub app is not installed" modal for non-admin repos authorized via /web-setup
- [BUG] Task list (TaskCreate/TaskList) not restored on --resume/--continue — task-list id resolves to a new runtime id
- Claude Desktop: Download SVG/PNG w widgecie (visualize/show_widget) nic nie robi
- [FEATURE] Multi-dimensional / scientific data visualization in Claude Code (charts, plots, heatmaps — beyond Mermaid)
- [BUG] Persistent "Unable to connect to API (ConnectionRefused)" on macOS
- [Bug] Excessive security alerts during normal SonarQube code review configuration
- [Bug] Excessive security alerts during normal SonarQube code review configuration
- permissions.ask does not fire for Bash commands under sandbox.autoAllowBashIfSandboxed: true
- Cloud routine environments silently changed AWS_* env handling twice (stripped, then platform-injected), breaking production workloads
- Remote Control: mobile session list shows stale content while a CLI session is actively running (recurring — see #31487, #30922, both auto-closed as dupes without a fix)
- [Bug] Excessive token usage in sessions after 5-hour reset period
- Desktop: MCP tool calls dispatched ~3-6 s apart while the server answers in 50 ms (CLI sessions unaffected)
- [BUG] IME input while scrolled up does not restore viewport to prompt input area
- Command scanner surfaces archived subdirectories (e.g. commands/_archive/) as namespaced commands
- [BUG]
- [BUG] デスクトップのClaudeアプリケーションが起動しない
- [Feature Request] Tool use approval prompts require modifier key to prevent accidental activation
- [Bug] Auto-invocation of Opus agent prevents manual code review mode
- [Bug] Input safeguard over-flags authorized red-teaming work, forcing fallback to Opus model
- Cloud tasks from deleted repo re-provision containers and show 'Session resumed' without user action
- [FEATURE] Tell the statusLine which subagent is focused when viewing its transcript
- [BUG] Claude Desktop — Code tab: "trust verification could not be completed" on folder select (silent failure, regression fixed by downgrade)
- [Bug] Network/Firewall Blocking Claude Code During Legitimate Research
- Opus (claude-opus-4-8, long context) confabulates tool-execution state — fabricated success reports and a fabricated prompt-injection/refusal, same env as #76764
- Headless `-p --permission-mode dontAsk`: denied tool → run reports success (exit 0, is_error:false, subtype:"success")
- Memory-file Write/Edit destroys all frontmatter when YAML parse fails (originSessionId stamper writes empty stub instead of rejecting)
- [BUG] Changing system time bugs chat history
- LSP tool: first query in a session returns incomplete results (cold-index race with server's initial workspace indexing)
- [Bug] Unable to update Claude Code
- [BUG] 2.1.206 Turning off Usage credits option returns completed but half finished workflows when workflows hit 5-hour usage limit
- Assistant response contained a fabricated user message appended after its reply; UI rendered it as a real user turn and harness re-fed it as human input
- I appreciate you reaching out, but I'm not able to process this request. This message appears to be in Chinese and discusses system access via OSCP, which doesn't relate to Claude Code (Anthropic's CLI tool) or a bug report that needs a GitHub issue title.
- Input during /compact silently cancels compaction; no machine-readable compaction state (multi-session orchestration)
- [BUG] PreToolUse hook `permissionDecision: "ask"` is overridden by `auto` mode for benign `dangerouslyDisableSandbox` commands (contradicts documented hook precedence)
- [Bug] Anthropic API Error: Intermittent failures occurring every ~5 messages
- Built-in guard against Claude reading/printing credential-file contents (.env, shared.env, etc.)
- Orphaned advisor_tool_result after auto-compaction breaks session with API 400 (still on v2.1.207)
- [Bug] Text before tool calls in same turn not displayed to user
- MCP HTTP client fails to resolve $defs/$ref in tool schema ("can't resolve reference")
- plugin marketplace update fetches new version but doesn't update installed_plugins.json
- VS Code extension: inline usage indicator shows stale percentage vs Account & Usage panel (93% vs 100% for same 5hr session limit)
- Repeated "verified" fixes fail because verification shares the bug's blind spot (Windows taskbar icon case study)
- Usage stats heatmap credits all activity to the session start date; days worked in long-running sessions show as zero
- Windows desktop app: projects, memory, and sandboxing are invisible or missing - field report from shipping a real product through it
- [Feature Request] Remove assumption of claude.ai hosting for generated HTML/CSS/JS artifacts
- [Feature Request] Auto-compact during background task idle with user-defined token threshold
- CLI exits silently when background-task notifications arrive (v2.1.207, Windows 11) - 4x in one day in a background-task-heavy project
- [Bug] Safety filter incorrectly flagging benign requests as harmful
- [Bug] Model downgrade to Claude 3.5 Sonnet during security patch implementation
- Runaway duplicate turns reread ~500k-token context and exhausted 5-hour usage limit in minutes
- [Bug] Model downgrade to Opus 4.8 triggered by guardrails during Fable usage
- [BUG] Cowork Project "Add folder" button permanently greyed out on Windows
- [Feature Request] Expand safeguard constraints for improved usability
- iOS web app paper cuts: input auto-zoom and sticky :hover on touch
- PreToolUse hooks stop firing after EnterWorktree switches session into a linked worktree (Windows, v2.1.207); /hooks reload doesn't fix it
- Mobile/web session list: surface hostname (and cwd) — same-repo multi-machine fleets are indistinguishable
- [BUG] `attribution.sessionUrl` should default to `false` (opt-in)
- [BUG] Mobile push notifications are never revoked after the session is answered on another device
- Workflow tool: StructuredOutput retry exhaustion silently accepts degenerate placeholder that passes loose schema — no retry/degraded signal exposed
- [BUG] fullscreen TUI copies via OSC 52 only, bypassing the native pbcopy path — silent clipboard failure in Terminal.app (local, non-SSH)
- Browser pane blocks shopee.co.id with "blocked by policy"
- [BUG] Character corruption in claude terminal - Linux Xfce4-terminal, calude version 2.1.207
- macOS Keychain: concurrent sessions race on OAuth refresh; setup-token workaround strips claude.ai connectors, leaving no viable headless pattern
- Workflow tool: args passed as a JSON array reaches the script as a JSON-encoded string
- Workflow resume: prefix-based cache re-runs all successful agents after an early failure (docs imply per-call caching)
- [MODEL] Fable5 - Silent downgrade to Opus
- LSP tool refuses to run a working, PATH-resolved clangd on Windows ("unsafe location (current directory)")
- [BUG] Japanese VM process name shows as mojibake in Activity Monitor (UTF-8 decoded as Shift-JIS) — still in 1.20186.1
- Native TaskCreate/TaskUpdate/TaskList/TaskGet become permanently unavailable mid-session, survives /clear and a full process restart
- Repeated pre-tool-call filler text (recurring stray "court" token) accumulates in long sessions, degrading throughput
- [Bug] False positive safeguard trigger on academic literature research with repeated API calls
- [withdrawn]
- [BUG] Intermittent Bun segfault at address 0x3C on session start — "Claude Code process exited with code 3" (Windows 11, Desktop app)
- [Bug] Agent ignores user confirmation requests and proceeds without verification
- [Bug] Table rendering fails with embedded photo content
- Computer use: all synthetic clicks/keystrokes silently dropped on macOS; screenshots work, permissions verified good
- [Bug] Unintended model switching from Fable 5 to Opus 4.8
- VSCode extension: extended-thinking blocks have no working expand control and vanish entirely after response completes
- [Bug] Anthropic API Error: Excessive false positives on authorized security testing work
- [Bug] Excessive false positives in content policy filtering for international customs/trade queries
- I appreciate you sharing the important context about your research, but I should clarify that I'm Claude Code, Anthropic's CLI tool for coding assistance. I'm designed to generate GitHub issue titles for bug reports, not to handle policy questions or acces
- [BUG] Claude UI stuck
- [BUG] AddPackage HRESULT 0x80073D28 persists after full SFC/DISM repair and reboot — related to #61392
- [Bug] Malformed tool-call generation at high context window utilization
- [BUG] Fable and All Models weekly usage multiplied by 4 (maxing out my fable usage) when upgrading from max 5x to 20x
- Interrupting mid-turn silently drops already-completed work from active context, with no warning and confident denial on query
- [Bug] Safeguard Flag incorrectly triggered for aging-related hyaluronic acid queries
- [Bug] Anthropic API Error: Overly broad safeguard blocking legitimate security code reviews
- [BUG] Agents view: reopening a stopped background agent dispatches a blank `mode: prompt` worker — same session ID, intact transcript never loaded, original prompt re-injected
- [BUG] Remote Control clients only see built-in slash commands — user-level commands and skills from ~/.claude are never advertised
- [BUG]
- [Bug] Unexpected cybersecurity block when generating images
- I don't have access to the bug report details associated with that request ID. To generate an appropriate GitHub issue title, I would need you to provide the actual bug report content or description of the problem you're experiencing with Claude Code. Ple
- [BUG] Claude Desktop macOS: MCP server child process cannot obtain Calendar (EventKit) TCC permission — Reminders works, Calendar doesn't
- [Feature Request] Subagents spawned via Agent tool should inherit CLAUDE.md
- Agent gave false 'not hanging' reassurances while a script it wrote burned 15,861 API calls over 13 hours
- HTTP MCP OAuth: refresh token stored but never exercised, server drops to bootstrap tools when the access token expires (~12h)
- [BUG] Desktop: Silently fails to include external referenced files or ask permission
- [Bug] Claude hallucinating user messages not sent, then contradicting itself during self-audit
- [BUG] Background Bash tasks (run_in_background) killed by external SIGTERM mid-execution, only after a session runs long
- Title: Desktop app sidebar shows no sessions/projects, though local session data is intact and CLI can resume them
- I need more information to generate an appropriate GitHub issue title. Could you please provide the bug report or issue description for Claude Code?
- SendMessage name-based addressing silently misroutes with multiple same-type subagent instances (should require/prefer agentId)
- [Bug] Claude refuses to execute commands and work. Claude has for the past 8 hours failed almost every request
- Desktop/web app no muestra el historial de sesiones (/resume) que sí lista correctamente el CLI
- Add Slovak (sk-SK) language support for voice dictation
- [Bug] Rate limit error reported despite usage below 50% threshold
- [BUG] Desktop app Remote Control: slash-command autocomplete force-corrects /clear to nearest match instead of allowing literal input
- [Bug] Agent executes file modifications without user confirmation
- [FEATURE] Gmail connector: opt-in access to the Spam folder (e.g. includeSpam), matching the existing includeTrash
- [FEATURE]
- [BUG] Desktop: sessions stay bound to dead Remote Control owner — no local re-home, Try-again loops, auto-RC toggle doesn't detach, local bypassPermissions suppressed
- [BUG] Workflow resume leaves orphan 'started' journal markers, inflating /workflows pending counts
- skill-creator: read_text()/write_text() sem encoding=utf-8 corrompem conteúdo acentuado no Windows
- [Bug] OTel: skill_activated masks skill.name to "custom_skill" while api_request events and cost/token metrics export it verbatim (regression since ~2.1.126)
- [Bug] Prompt injection vulnerability detected in code execution context
- SendMessage to a completed subagent doesn't reuse the previous prompt prefix (cache miss on every resume)
- Desktop app: no discoverable way to switch project/working folder for a new session
- Local scheduled tasks: silent no-trace failures, stall-then-burst late firing, alive-but-stalled fires, and missed-run ledger desync (4 incidents in 8 days, macOS desktop app)
- [Bug] Credits system not functional
- Provide a structured DAG view for orchestrator/subagent workflows
- [BUG] Fable 5 → Opus 4.8 fallback is sticky for the rest of the session; benign planning/orchestration turns never return to Fable (manual /model is the only escape)
- Embedded Browser pane shows blank content even though the page loads fine
- Dispatch: add a "Log and Refresh Chat" button to checkpoint context before it gets foggy
- [Bug] Claude Code working in wrong directory, mixing project histories
- Cowork sessions can no longer manage local scheduled tasks: scheduled-tasks MCP not exposed to (now remote-executing) desktop sessions, and the Scheduled folder is blocked as a protected location
- [Bug] Systematic usability issue - unresponsive behavior
- [BUG] Agent tool spawns 7 named teammates for a single routine task; nested spawning burns 77% of 5h limit in <1 hour
- [Bug] Anthropic API: Requests Flagged as Unsafe on SaaS Platform
- Voice tap doesn't work in agent view
- CLAUDE_PLUGIN_ROOT path separators stripped on Windows-client + WSL-mounted project (semgrep@2.1.1 MCP fails 100%)
- [BUG] Background Bash tasks (run_in_background=true) are sporadically SIGKILLed mid-run by the CLI's task supervision (~1% of dispatches; can corrupt git state)
- No mechanism to require checkpoint/approval before large-scope generative rewrites of user content
- [BUG] Custom URL scheme links (message://) not opened by OS in Claude Desktop — same root cause as #26952
- [Bug] Anthropic API Error: Validation failures on Ansible playbook analysis
- [BUG] Gmail connector: get_thread silently truncates long threads — newest messages invisible to the agent (refiled from claude-ai-mcp#239)
- [Bug] Setup documentation unclear and confusing
- [BUG] Claude Desktop (Windows): Squirrel→MSIX migration deletes the old install but orphans existing shortcuts (desktop AND Start Menu) — confirmed on 2 machines
- Feature request: auto-return to plan mode after an approved plan finishes executing
- Cowork/Workspace EXDEV rename fails within the SAME directory (not just cross-Temp/Roaming) — MSIX install, v1.19367.0.0
- [Feature Request] Allow code review of user's own source code without flagging
- [Bug] Unexpected non-ASCII characters in Claude Code output
- Weekend post-mortem: fuck-all got built, and Fable still ate its usage on process it invented instead of the work I asked for
- [FEATURE] Re-run statusLine command on terminal resize (SIGWINCH)
- VS Code extension: prompt input focus ring ignores theme `focusBorder` (hardcoded orange)
- Bash permission `allow` rules (wildcard and exact-match) never suppress prompts on WSL project opened via `\\wsl$` UNC path (VSCode WSL Remote)
- [Bug] API key environment variable name triggers security filter
- [FEATURE] Voice input (microphone) does not support Russian / non-English speech recognition language selection
- [Bug] Unnecessary cybersecurity warnings for authenticated local development
- Windows: project entries in ~/.claude.json keyed case-sensitively - CLI and VSCode sessions silently see different config for the same folder
- [Feature Request] Auto-cleanup empty .claude/.cc-writes/ directories after execution
- Dispatch cannot start or message Code sessions — persistent session_stale_relogin
- [Bug] Anthropic API Error: Recurring/Repeated API Failures
- [Bug] Excessive API retry attempts consuming quota after limit restore
- [BUG] Agent per-invocation model schema rejects [1m] values accepted by subagent profiles
- Default Model for Scheduled Tasks
- Disproportionate subscription usage burn in long sessions: cache TTL expiry re-writes + no usage attribution in /usage
- [Bug] Anthropic API Error: Request blocked on simple arithmetic query
- [FEATURE] VS Code extension: show model, Ultracode/mode, effort, and usage indicators next to the chat input (desktop app parity)
- IDE extension: expanding a long user message pushes 'Show less' off-screen with no in-message scroll — can't collapse without scrolling the whole session
- [BUG] Cowork VM hangs mounting Plan9 share during startup — deterministic, host-side, distinct from #76094
- [Bug] Model automatically downgraded from Claude Fable 5 to Claude 3 Opus 4.8 due to pet food content analysis
- VS Code extension: assistant's text reply is hidden when the turn ends with ExitPlanMode (plan approval card)
- [BUG] Worktree isolation targets the outer repo, not the nested child repo the session started in
- Korean (Hangul) IME composition inserts stray characters (e.g. "?") across CLI TUI and Agents view
- queued messages and unsent input drafts are silently lost on session switch
- [Feature Request] Auto-save session with default name on each user interaction
- [BUG] Anti-cheat driver crashes bundled Bun runtime; surfaced only as "process exited with code 3"
- [Bug] Model downgrade triggered by "cdc" substring in column names
- Line breaks render as multiple blank lines in chat composer/message bubble (worsens with paste round-trips)
- Clicking a notification while a multiple-choice question is open cancels the question (acts like Escape)
Resolved issues (160)
- [BUG] AskUserQuestion preview pane renders blank when content starts with an HTML-block tag (XML / HTML / SVG / Vue SFC)
- [BUG] ConnectionRefused on /v1/messages streaming while simple API requests succeed (Windows, Bun)
- Hooks fail open when CWD is invalidated (e.g. worktree deletion)
- Allow intentional empty/whitespace-only assistant responses without the "no visible output" nudge
- Remote Control: mobile session list shows stale content while a CLI session is actively running (recurring — see #31487, #30922, both auto-closed as dupes without a fix)
- [Feature Request] Relax safeguard restrictions for advanced users
- [Bug] Anthropic API Error: Opus safeguards blocking tool use execution
- Model spontaneously generated a fake "prompt injection test" block at the start of an assistant turn
- [Bug] Dynamic Workflow blocked by Claude 4.8 Opus in auto mode only
- I don't have enough information to generate a GitHub issue title from this request ID alone. Please provide: 1. A description of the bug or feature request 2. Any error messages encountered 3. What you were trying to do when the issue occurred 4. Expected
- PreToolUse `updatedInput` from an `allow` hook is silently discarded when a sibling hook returns `ask` and the user approves
- [MODEL]
- [Bug] Request declined: Unable to process non-cybersecurity topics
- [Bug] Anthropic API Error: Content Policy Violation on Non-Sensitive Configuration Output
- [MODEL] sonnet 5
- [Bug] Tool output misrepresentation causing inaccurate behavior reporting
- [FEATURE] /resume should sort by last conversation activity, not last-touched — and expose a persisted sort preference
- [Bug] Tool rendering fails after extended use, causing cascading failures
- [BUG] Cowork silently moved a long-running local agent to the cloud, and cloud mode breaks its workflow
- Remote Control not enabled for Pro account
- [BUG] CLAUDE_CONFIG_DIR does not redirect global CLAUDE.md — still loaded from hardcoded ~/.claude/CLAUDE.md
- 検証フィードバック：業務データ登録タスクにおける探索精度とツール制約について
- [Bug] Security verification incorrectly flagged during pre-release product check
- [Bug] Skills no longer visible in CLI interface
- [Bug] Opus context window limited to ~500K tokens despite 1M advertised capacity
- [Bug] Unwanted Wordpress interference blocking file editing
- I don't see a bug report in your message. Please provide details about the issue you're experiencing with Claude Code, including: - What you were trying to do - What error or unexpected behavior occurred - Any error messages or logs - Your system informat
- [Bug Report: Repeated errors in Fable 5 project compilation
- I appreciate you providing context, but I don't see a specific bug report or technical issue to create a GitHub issue title for. To help you generate an appropriate issue title, please provide: 1. **What problem occurred?** (e.g., error message, unexpec
- [FEATURE] Wire `sandbox.allowPty` from settings.json into the macOS Seatbelt profile (sandbox-runtime already supports it)
- [BUG] Claude Desktop Code: Auto-Reconnects to Last SSH Host on Startup
- [Bug] Claude outputs duplicate instructions and repeats text on terminal resize/scroll
- Desktop app ignores `model` field in ~/.claude/settings.json
- VS Code extension: terminal-mode Claude not respawned after window reload (panel returns empty)
- Conversation content duplicates in scrollback after terminal resize (font-size / window-size)
- [FEATURE] visibility: "team" for routine-created sessions (config-level + /fire override)
- [BUG] Headless mode (`-p --output-format stream-json`) breaks AskUserQuestion handling — model continues on assumptions instead of stopping
- [Bug] Anthropic API Error: Usage Policy Violation Without Clear Cause
- [Feature Request] Add ability to resume/continue stalled deep-research workflow runs
- [BUG] Claude Code executed destructive command (adb shell pm clear) against explicit user instruction, causing permanent data loss
- [BUG] Session terminates automatically when using multiple subagents.
- [Bug] Chrome native host crashes with ERR_INVALID_ARG_TYPE on Windows named pipe creation
- test
- [BUG] False-positive AUP block on legitimate comparative-genomics annotation (Request ID included)
- False-positive Usage Policy blocks on legitimate academic sessions: safety classifier fires mid-session on accumulated biomedical vocabulary
- AUP false positive on medical imaging project (intracranial hemorrhage detection)
- [BUG] says "Server is temporarily limiting requests"
- [MODEL] Opus 4.6 - I told it to skip other data and it still imported other data.
- [Bug] Model classifier rejects claude-opus-4-7 as invalid model
- Auto-compact circuit breaker prevents compaction on Haiku (v2.1.65+)
- Agent teams mode: touch scroll broken over RDP (works in regular mode, same terminal)
- [FEATURE] Generic bindable action to send literal text / run a command from a keypress
- Proposal: Behavior-memory budget advisor (Loop Pilot) as pre-loop hook
- API Streaming Stalls Causing 59-138 Second Delays and Timeouts
- Plugin MCP reconnect fails: ${CLAUDE_PLUGIN_ROOT} not substituted after /model or session reconnect
- [BUG] Claude Max 5x: 1M context window auto-triggered in Cowork session, deducting usage credits without user request
- [MODEL] opus 4.8
- [Bug] Anthropic API Error: Security Policy Violation on Legitimate Code Project
- [Bug] AWK variable syntax ($2) stripped from skill definitions
- [BUG] Silent model switch to Opus 4.7 [1M] mid-session caused ~4× quota burn
- [BUG] /ultrareview run failed
- [Bug] Ultrareview crashes during dedupe step without delivering final report
- [BUG]
- [Bug] Anthropic API Error: False-positive Usage Policy block on biomedical research with Opus 4.8
- [DOCS]
- [BUG] Branch selector missing in iOS Code UI
- [BUG] `claude.workspaceRoot` changes context but not session list scope
- Plugin subagents can't resolve ${CLAUDE_PLUGIN_ROOT}/${CLAUDE_PROJECT_DIR} — no way to read plugin-bundled files from a subagent
- MCP server registered via 'claude mcp add' not surfacing tools in conversational interface when claude.ai web connectors are also present
- Docs: plugin MCP tool namespacing + allowed-tools vs subagent tools: enforcement are under-documented
- Stats dashboard counts multi-day sessions only on their start date (Active days / streaks / peak hour undercounted)
- [FEATURE] Session Teams (Make structured sessions and sessions can comunicate each others interactively)
- [Feature Request] Opt-out for "Feature of the week:" splash line
- [Bug] Approval prompt fails to render intermittently, causing session to hang
- [BUG] Frequent session freezes / connection drops on Claude Code (VS Code extension) under
- Remote Control: Android approval signal silently dropped — server never receives it (Linux/tmux, pure remote session)
- Spurious "Your tool call was malformed and could not be parsed" injected after tool calls that actually executed
- V8 OOM: Claude Code CLI leaks memory until laptop crashes (113GB observed)
- [BUG] I am eligible for 2x more usage until July 5, but it doesn't apply on my usage page
- [BUG] 'No Json object found in response' when creating an agent using '/agents' command
- [BUG] Allow partial globs in permissions.allow for MCP tools (regression 2.1.166+)
- Remote Control fails: `bridge_state: failed - /login` (bridge reachable, valid personal-account token rejected)
- [Bug] Anthropic API Error: Usage Policy violation on legitimate code audit request
- [withdrawn]
- VSCode extension: extended-thinking blocks have no working expand control and vanish entirely after response completes
- LSP bridge: goToDefinition/findReferences/workspaceSymbol/goToImplementation empty while hover/documentSymbol work — call hierarchy result inconsistent across sessions (independent repro of #44767 / #32499)
- I need more information to generate an appropriate GitHub issue title. Could you please provide the bug report or issue description for Claude Code?
- [BUG] Bug Report: Claude Desktop for Windows — Update fails to register due to stale background processes ("still running" deferred-registration loop)
- [BUG] Billing: ~€85 promotional credit disappeared without usage (Max 5x, April 2026 promo)
- claude desktop: press F2 to rename the hovered session in the sidebar list
- [BUG] Claude Desktop MSIX install fails with HRESULT 0x80073CFF despite full elevation, sideloading enabled, and clean WindowsApps\Deleted folder
- [FEATURE] Programmatic way to file a session into a Claude Desktop Project
- [Bug] Model hallucinates and invents content instead of reading files accurately
- [Bug] Tool refuses to execute task for publicly available content
- tool deosnt follow clone- FABLE SHIT
- [Bug] Claude Code reverting changes unexpectedly and losing work context
- [Bug] Model downgrade during bug fixes causing degraded performance
- [Bug] Claude Code fails to handle Expo CLI commands
- [Bug] Anthropic API Error: Missing safety violation details in flagged message response
- [Bug] Claude Code always falls back to Opus for any task
- I need more specific information to create an accurate GitHub issue title. Could you provide: 1. What is the flag that's wrong? 2. What is "Fable" in this context (a feature, module, component)? 3. What is the incorrect behavior vs. expected behavior? 4.
- [Bug] API Rate Limiting or Model Access Restrictions Preventing Legitimate Use Cases
- [Feature Request] Usage credit adjustment for Anthropic API errors and rate limiting issues
- `/desktop` fails with "CLI session transcript not found" when run as the first command in a session
- `/desktop` session transfer drops the EnterWorktree working directory — Desktop resumes in the main checkout (wrong branch), no mismatch warning
- [FEATURE] Desktop: expose worktree name/base input in UI + pass branch-prefix & base branch into the WorktreeCreate hook payload
- [BUG] /desktop "cannot determine working directory" after EnterWorktree→ExitWorktree — session transcript fragmented across two project slugs; /desktop resolves to a bridge-session record with no cwd
- [BUG] Desktop branch/worktree indicator not updated after EnterWorktree — shows main-checkout branch while session cwd is correctly in the worktree
- [BUG] Day confusion
- [DESIGN] .gitignore should not double as AI access control — introduce clear .claudeignore semantics
- [BUG] [BUG] Cowork workspace fails with false "Download failed" on Windows MSIX 1.11187.4 — VM files validate OK but download_and_sdk_prepare throws before smol-bin copy step
- [BUG]
- [BUG] Claude Dispatch permanently showing "desktop appears offline" for multiple weeks
- Feature: Hook output type `attachFile` to inject large content without size cap or terminal noise
- [FEATURE] Slash-command autocomplete dropdown in claude.ai/code web UI (parity with the desktop app)
- [FEATURE] Show file path in image placeholder when context is compressed
- [BUG] Network Sandbox in WSL intermittent failures
- Dispatch feature ignores CLAUDE_CODE_DISABLE_1M_CONTEXT — "Usage credits required for 1M context" blocks background agents
- [BUG] Plugin directory does not render in Claude Desktop
- [FEATURE] Default deny permissions for *.env files
- [Bug] Usage credits required for 1M context on Max 5x plan — standard model selected, Cowork + Claude Code
- Custom theme overrides ignored in startup banner title (borderText pre-coloured with base theme)
- Session sync between Desktop app and VS Code/Cursor extension
- Feature: Allow existing Claude sessions to communicate with each other
- Show status line during permission prompts and user questions
- [Bug] Opus 4.8 responses not displayed after tool use — quota consumed with no output
- Native installer leaves 0-byte stub on interrupted update, breaks launcher
- [Bug] Image preview fails with error icon in remote control mode on mobile
- Office 365 MCP connector: add send email capability (currently read-only)
- [BUG] limit reached
- [BUG] WorktreeCreate hook stdin payload doesn't match documentation — Workflow runtime sends 'name', docs say 'worktree_name' + 'base_path' + 'source_ref'
- Proposal: deprecate Cowork and consolidate its capabilities into Claude Code
- [BUG] Managed MCP server tools permanently stuck in "Needs Approval" — users cannot approve unlisted tools, contradicting docs and UI
- [BUG]
- Desktop App: OAuth token refresh kills all active Claude Code sessions with exit code 143 (SIGTERM)
- [FEATURE] Cowork: Make "Allow once" easier to select in MCP permission dialogs
- Opus 4.8 + dynamic tool loading (ToolSearch) re-modifies signed thinking blocks → "messages.N.content.M ... cannot be modified" 400 storm → strip-and-retry every turn
- [BUG] VS Code model picker lists Sonnet 4.6 twice (alias + pinned ID)
- claude install adds ~/.local/bin to ~/.bashrc only — macOS login shells do not read it, so `claude` is missing from PATH in new terminals
- Silent paste truncation on Windows — no warning when input is clipped
- [BUG] PowerShell tool fails Exit 1, empty stdout/stderr on Windows 10 DE-locale (v2.1.142, all environmental causes ruled out)
- Feature Request: Add /messages command to browse conversation history
- [BUG] Bypass Permissions warning silently persists acceptance forever — prompt says "while running in … mode" but the choice is permanent
- [FEATURE] Feature Request: UI localization support (e.g., Portuguese / pt-BR)##
- [Bug] AUP Classifier False Positives on Legitimate RMM/Remote Desktop Vocabulary
- Post-compaction: stale skill ARGUMENTS re-surfaced as fresh prompt; last user message dropped from summary
- [Bug] Chrome extension pairing requires re-selection after container rebuild despite valid persisted device ID
- [BUG] Transcribing that should have been reproduced verbatim
- [BUG] Claude desktop fails to launch after applyin an update
- [BUG] 1M context required error on active session — Max plan, Sonnet 4.6 High Effort, v2.1.157
- [BUG] Monitor usage creates fake Subagent completion events
- [FEATURE] Agent view should not auto-complete sessions — require manual completion or archive
- [Feature Request] Add model version pinning to ensure reproducible behavior across Anthropic API updates
- [BUG] "Usage credits required for 1M context" error when context is set to Medium on Sonnet 4.6
- [Bug] Anthropic API Error: Usage Policy Violation - Cyber Safeguards Triggered
- False-positive safety/AUP interruptions during legitimate trading-app development (testnet, own project)
- [Bug] Cybersecurity protection blocks Claude model execution during code writing
- [Bug] Anthropic API Error: Usage Policy violation on git push operations
- Claude recommends Flask catch-all static routes without securing against sensitive file exposure (.env, .git, .py)
- [Feature Request] Auto-save session with default name on each user interaction