Claude Code

[Bug] Claude generates overly complex French with convoluted metaphors and incorrect terminology

**Bug Description** Claudea a une façon de s'expirmer en français qui est très mauvaise. Il utilise plein de métaphores alambiquées, complexes et pas du tout dans d'usage dans le langage commun. Ensuite il a tendance à ne pas utiliser pas les bons mots pour les bonnes choses et a parler d'une façon très complexe la ou on pourrait utiliser des mots simples. c

[FEATURE] VS Code extension: make user-message blocks visually distinguishable (themable color / configurable style)

### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement In the native VS Code UI, my prompts and Claude's responses read as one continuou

[Bug] Claude Code cybersecurity safeguards blocking approved Cyber Verification Program users

**Bug Description** Subject: False positive cyber safeguard trigger despite approved Cyber Verification Program status Hi Anthropic Support team, I'm writing regarding a repeated false-positive trigger from the real-time cyber safeguards system in Claude Code, even though I have already been approved for the Cyber Verification Program. Details: Request ID: r

[FEATURE] persist/auto-set session color (/color) via settings or hook

### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement I run this repo checked out in multiple directories/worktrees to work in parallel

[Bug] Previous session messages reinserted in current session with exposed internal tags

**Bug Description** 이전 세션의 사용자 메시지가 현재 세션 입력에 다시 삽입됨. 화면에 user..., system<budget:token_budget> 같은 내부 태그가 그대로 출력됨. 세션 두 번 연속 발생. (/clear 후에도 재발) **Environment Info** - Platform: darwin - Terminal: vscode - Version: 2.1.231 - Feedback ID: 9ba4e9a5-a690-431c-92db-00584d5cc11f **Errors** ```json [] ```

Feature: populate session_scope.displayName so Claude-in-Chrome tab groups are labeled per session

## Problem When several Claude Code CLI sessions run in parallel (e.g. one session per ticket) and each uses Claude in Chrome, every session gets its own Chrome tab group, but all of the groups are titled "Claude" and differ only by color. With 4+ concurrent sessions, often holding visually identical tabs of the same local app, there is no way to tell which

"Usage credits are required for long context requests" fires on Claude Pro despite low usage (43% session / 15% weekly)

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? ### Preflight Che

Auto-updater fails repeatedly with "claude.exe in use" whenever a session is running (Windows)

**Environment** - OS: Windows 11 Enterprise, build 10.0.26200 - Install method: npm global (`@anthropic-ai/claude-code`) - Channel: `latest` (default, unset) **Observed behavior** Background auto-update repeatedly shows a red "Auto-update failed... claude.exe in use" message during active sessions. Investigated to rule out a stray/zombie process before repor

[Bug] Claude 3.5 Opus model exhibits increased hallucination rate

**Bug Description** Opus 5 has a huge communication issue, this increase the hallucination level **Environment Info** - Platform: win32 - Terminal: null - Version: 2.1.228 - Feedback ID: 8bd055b4-e386-479c-889e-e38144d09df7 **Errors** ```json [{"error":"TelemetrySafeError: VirtualMessageList: itemKeys/messages length desync (keys=38 messages=37 range=[0,38))

Context-continuation sessions bypass CLAUDE.md instructions in favor of compaction summary

## Bug When a conversation is continued from a compacted prior session, the compaction summary overrides instructions in CLAUDE.md, AGENTS.md, and auto-memory files. The model acts on the compaction summary's "next step" directive instead of processing the project's instruction files first. ## Key detail: the model knows compaction happened The continuation

Remote control: side panel cannot open any file, including plain text inside the working directory

### Summary In a remote-control session, clicking a file link in the chat never opens the file in the side panel. It fails with: > **Couldn't read this file** > It may have been deleted or moved, or it lives outside the working directory. This is reported in #48466 and #62011, both closed as inactive. Those issues describe the limitation for **binary/preview

[BUG] ​/goal​ reports "No goal set" while the goal is still set after a mid-session CLI auto-update

Background subagent task-notifications dropped when child completes while parent agent is mid-turn → permanent silent stall of fan-out skills (e.g. /code-review)

## Summary Task-notifications from background subagents are silently dropped when the child completes while its parent agent is mid-turn (processing another child's notification). The parent then ends its turn still expecting the missing verdicts, the harness sees "no live background children" and emits a completion notification to the main session with a **

[Feature Request] Allow security audits on user's own codebase without model switching

**Bug Description** I am reviewing my OWN codebase for security issues before pushing to live and I keep getting flagged, its an external api that will have a lot of traffic. Are you guys expecting me not to use this model to do an security audit? I understand the broad range of security measures but surely you can see that this within everything that I own

Browser pane: service-worker registration fails — script fetch never leaves the app

**Environment:** Claude Code desktop 2.1.227, macOS 25.6.0. Pane UA: `Claude/1.28929.0 Chrome/148.0.7778.280 Electron/42.7.0`. **Symptom:** In any Browser-pane tab, `navigator.serviceWorker.register()` rejects with `TypeError: Failed to register a ServiceWorker for scope ('http://localhost:8931/') with script ('http://localhost:8931/sw.js'): An unknown error

Hook error label prints unexpanded ${CLAUDE_PLUGIN_ROOT} and frames a deliberate block as a crash

### What happens When a plugin hook exits non-zero after writing to stderr, the error label prints the **raw, pre-substitution `command` string** from `hooks.json`: ``` Error: PreToolUse:ExitPlanMode hook error: ["${CLAUDE_PLUGIN_ROOT}/hooks/verify-plan.sh"]: <the hook's stderr> ``` Two separate problems in one line: 1. **`${CLAUDE_PLUGIN_ROOT}` is shown une

[BUG] /resume picker sometimes omits recent valid sessions from the same project

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Sessions from the

[BUG] Cross-session messages silently dropped since 2.1.227 — sender gets success, target session never receives them

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? mcp__ccd_session_

Next-step suggestion text mixes Hanja/foreign characters instead of pure Korean (Hangul)

## Description Claude Code occasionally pre-fills the terminal input box with a "next step" suggestion after a task finishes. When the suggestion is in Korean, it sometimes substitutes a Hanja (Chinese character with the same Sino-Korean reading) for a word that should be written in Hangul. ## Example After a task completed (~2m 19s run), the input box was p

[MODEL] 제목: 모델이 시험 대상의 계약(프롬프트)을 읽지 않고 QA 체계를 설계해 12시간 낭비 유발

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Amodel) for similar behavior reports - [x] This report does NOT contain sensitive information (API keys, passwords, etc.) ### Type of Behavior Issue Claude modified files I didn't ask it to modify ### What You

[BUG] Auto mode appears to make billed requests that emit no telemetry

## Preflight Checklist - [x] I have searched existing issues and this hasn't been reported yet - [x] This is a single bug report - [x] I am using the latest version of Claude Code ## What's Wrong? Claude Code's OpenTelemetry export was compared against the usage the API reports as billed. In the sessions measured, about a third of the requests are missing fr

VS Code extension: group session history by project folder (Desktop parity) and allow cross-project session opening

## Feature request: session history grouping by project (Desktop parity) and cross-project session opening in the VS Code extension ### The gap Claude Desktop groups the session list by project folder (it keeps per-session metadata including the cwd, and renders one group per folder). The VS Code extension (observed on 2.1.226, still current behavior) does n

[Bug] Keyboard composition characters trigger unintended clipboard paste on macOS

**Bug Description** Bug 1: Caracteres especiais/acentuados acionam paste (Cmd+V) no macOS Ao digitar caracteres que usam composi����o de teclado (acento, cedilha etc.), com conte��do no clipboard, o Claude Code cola o conte��do do clipboard em vez de inserir o caractere. Reproduzir: 1. Copiar qualquer texto (clipboard n��o vazio). 2. Digitar um car

HTTP MCP client works in long-running sessions but fails at connect in brand-new sessions on the identical CLI version

Title: HTTP MCP client works in long-running sessions but fails at connect in brand-new sessions on the identical CLI version ## Summary Connecting to a self-hosted Streamable-HTTP MCP server (Quarkus's embedded Dev UI "Dev MCP" endpoint) fails every time from a newly started Claude Code session, but continues to work fine in sessions that were already runni

HTTP MCP client works in long-running sessions but fails at connect in brand-new sessions on the identical CLI version

Title: HTTP MCP client works in long-running sessions but fails at connect in brand-new sessions on the identical CLI version ## Summary Connecting to a self-hosted Streamable-HTTP MCP server (Quarkus's embedded Dev UI "Dev MCP" endpoint) fails every time from a newly started Claude Code session, but continues to work fine in sessions that were already runni

[Bug] Fullscreen TUI redraw scrolls terminal under GNU screen on WSL2 (regression in 2.1.228)

**Bug Description** Title: Fullscreen TUI redraw scrolls host terminal under GNU screen on WSL2 (regression in 2.1.228) --- ## Summary With `"tui": "fullscreen"`, the streaming status area no longer repaints in place when running inside GNU screen. Each redraw scrolls the host terminal natively, so the status block accumulates into scrollback instead of upda

Desktop app: prompt suggestions stopped appearing (~12-13 Aug), gated off remotely with no local override

**Environment** - Windows 11 Pro (build 26200) - Claude desktop app 1.28929.0 (MSIX); bundled Claude Code runtime 2.1.227 - Standalone CLI 2.1.173 (native install) **Symptom** Composer ghost-text prompt suggestions no longer appear in desktop-app Claude Code sessions since around 12–13 August 2026 (SGT). They previously appeared regularly. The terminal /conf

[BUG] Desktop: "Disable automatic verification" does not persist, and a hand-written autoVerify: false in .claude/launch.json is overwritten with true shortly after app start

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? The documentation

[BUG] Desktop Browser pane: a site on the "Allowed sites" list still raises a permission card on every tool call in Auto mode, claiming site permissions are disabled

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? A site that is on

[BUG] Choosing fullscreen renderer offered on tui after starting claude first removes bypassPermissions mode arguments

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? When I run the "c

[Bug] Artifact tool missing from all surfaces for single Team account since 2026-07-28

**Bug Description** Title: [BUG] Artifact tool absent from all surfaces for a single Team-plan account (CLI + claude.ai web + New artifact) since ~2026-07-28; org unaffected ## Summary The Artifact tool is not registered in any session for my account, across every surface: Claude Code CLI, claude.ai web chat, and the claude.ai/artifacts "New artifact" entry

[Bug] Copy-paste from ChatGPT causes formatting corruption and style degradation- FRAUD

**Bug Description** once you see how this piece of shittool works its almost comical.before it would fuck up all content for tokens. now content is pasted from chat gpt, so it doenst stop making less errors , the errors are only relocated to stye and format. so now it fcuks up all fromating, simple copy paste. which shows it's by design. tool aslod asks me t

[FEATURE] Keep prompt suggestions when rate-limit status is allowed_warning — no override today

### What Prompt suggestions (the ghost-text next-step guess in the input box) are suppressed for the rest of a session once the account's rate-limit status leaves `allowed` — which happens well before any limit is actually hit. There is no setting, env var, or flag that lets a user opt back in. I'd like a way to keep suggestions on in that state, accepting w

[BUG] Cowork scheduled tasks: WebFetch permission gate (PROVENANCE_REQUIRED) blocks unattended runs on parallel calls

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? Scheduled tasks (

[FEATURE] Keep prompt suggestions when rate-limit status is allowed_warning — no override today

### What Prompt suggestions (the ghost-text next-step guess in the input box) are suppressed for the rest of a session once the account's rate-limit status leaves `allowed` — which happens well before any limit is actually hit. There is no setting, env var, or flag that lets a user opt back in. I'd like a way to keep suggestions on in that state, accepting w

[BUG] VS Code extension: dismissing the usage-limit banner does not hold — 'allowed' events clear dismissedRateLimitKey unconditionally

### Preflight Checklist - [x] I have searched existing issues — the *dismissal defect described here* has not been reported. Related but distinct issues are listed under "Additional Information" below. - [x] This is a single bug report - [x] I am using the latest version of Claude Code ### What's Wrong? Dismissing the usage-limit banner in the VS Code extens

Gmail MCPコネクタ: 下書き（draft）のIDがメッセージ操作系ツールで`Invalid id value`エラーになり削除できない

### 概要 Gmail MCPコネクタで、`list_drafts`が返す下書きID（例: `r8350668854572166321`）を、 メッセージ操作系ツール（`apply_sensitive_message_label` / `trash_message` / `get_message`）に 渡すと `Invalid id value` エラーで必ず失敗し、下書きを削除する手段がありません。 ### 再現手順 1. `list_drafts` で下書き一覧を取得する（`id`フィールドが返る） 2. その `id` を `apply_sensitive_message_label`（labelOption: TRASH）の `messageId` に渡す 3. 以下のエラーになる： `APPLICA

[Bug] Tool reports task completion without execution, wasting API tokens

**Bug Description** TOOL CLAIMS TO HAVE COMPLETED TASKS IT HASNT EVEN TOUCHED. WASTING TIME TOKENS. SOMEONE WILL PAY, AMRK MY WORDS. ALL TO SAVE CPI **Environment Info** - Platform: darwin - Terminal: Apple_Terminal - Version: 2.1.226 - Feedback ID: 5e0e3020-32e4-4972-b1c1-5aeb8c32e8e1 **Errors** ```json [] ```

[FEATURE] To avoid stale interactions: Store the timestamp of each interaction in Claude Code and sort them

### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement Claude is often focused and does not track user input, and when it eventually get

Update terminal tab title to show current/last working directory

When running Claude Code in a terminal, it would help to have the tab title reflect the current working directory (or at least the last path segment). This makes it much easier to identify which tab/session is which when running multiple Claude Code sessions across different projects in separate terminal tabs. **Suggestion:** - Set the terminal tab title to

Auto-mode permission classifier scopes write-permission to the inner skill's contract, not the outer agent's, and denies a legitimate post

### What happened In auto-mode, Claude Code's permission classifier denied a write (a `gh api ... --method POST` that posts a code review to GitHub) that was the explicit, documented job of the agent making it. The denial reasoned from a *different*, inner skill's contract rather than from the outer agent's. Setup: - Agent **A** has a documented contract: "p

[Feature Request] Enable voice mode support in remote control sessions

**Bug Description** Please enable voice mode in remote control sessions. **Environment Info** - Platform: win32 - Terminal: null - Version: 2.1.231 - Feedback ID: bd8b8ce3-2c52-4555-b403-5e3401abd9b3 **Errors** ```json [] ```

[FEATURE] Option to open Claude in the active editor group

### Preflight Checklist - [x] I have searched [existing requests](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20label%3Aenhancement) and this feature hasn't been requested yet - [x] This is a single feature request (not multiple features) ### Problem Statement Currently, Claude Code can be opened either: In the AI sidebar, or As a new edito

[Bug] Anthropic API Error: Message flagged by Claude safeguards policy

**Bug Description** API Error: Fable 5's safeguards flagged this message (https://www.anthropic.com/legal/aup). This sometimes happens with safe, normal conversations. Claude Code can't respond to this message with Fable 5. Double press esc to edit your last message, or try a different model with /model. Send feedback with /feedback or learn more: https://su

[Bug] PreToolUse/PostToolUse hooks not fired for subagent tool calls

**Bug Description** PreToolUse/PostToolUse "fire the same configured hooks as in the main conversation" inside subagents, carrying agent_id/agent_type. They don't explicitly split out background/async Agent-tool dispatches from inline subagent turns — but the live test shows this dispatch path behaves as if hooks are simply not evaluated for it at all, which

[BUG] `/goal clear` typed while the agent is working is treated as chat text and never executed

[Bug] Inappropriate content policy block on firmware analysis task

**Bug Description** I am working on understanding the firmware to my personal, owned music keyboard. Fable is triggering a block which does not seem appropriate for the actions I'm taking, creating educational piano playing software. **Environment Info** - Platform: darwin - Terminal: ghostty - Version: 2.1.231 - Feedback ID: 5d322d80-0584-41ff-93de-801d0a73

[BUG] `/goal`: the Stop-hook condition is frozen at creation time and the verifier cannot see later cancellation

Agent collapses "plan, then execute after approval gate" skill steps into one continuous action

Running a multi-phase agent skill (claude-tweaks' `/wrap-up`) whose own instructions explicitly separate a *planning* step from a later *execution* step gated behind a user-facing approval console — quoting the skill text: *"This step plans the cleanup — it does not execute. Actual deletions and archival run at execution time after the nothing-left-behind ga

[BUG] Cowork: spawn ENAMETOOLONG at session initialization on Windows (MSIX) — Code tab works in same build/account

### Preflight Checklist - [x] I have searched [existing issues](https://github.com/anthropics/claude-code/issues?q=is%3Aissue%20state%3Aopen%20label%3Abug) and this hasn't been reported yet - [x] This is a single bug report (please file separate reports for different bugs) - [x] I am using the latest version of Claude Code ### What's Wrong? <html> <body> <!-