Digest / June 9, 2026
Claude Code Issues Digest — June 9, 2026
343 new issues filed ·
194 resolved in anthropics/claude-code (UTC day).
New issues (343)
- System prompt hardcodes stale user path (-home-e) and writes to disk without user consent
- [BUG] Claude Code (VS Code ▎ extension) — copying a code block ▎ injects newlines into long commands
- [FEATURE] Add a CLI command to hand off a running local session to the cloud (local→web, the inverse of `--teleport`)
- [BUG][Windows][native-binary] v2.1.159+ silent-exits during startup — v2.1.112 unaffected
- `/deep-research`: `[verified]` tags aren't bound to a primary-source check — checkable facts can pass on inter-model agreement / secondary sources
- [DOCS] `--safe-mode` flag and `CLAUDE_CODE_SAFE_MODE` env var undocumented in CLI reference, env-vars, and troubleshooting pages
- [DOCS] `/cd` command missing from the built-in commands reference table
- [DOCS] Settings and env-vars pages omit `disableBundledSkills` setting and `CLAUDE_CODE_DISABLE_BUNDLED_SKILLS` environment variable added in v2.1.169
- [DOCS] [Interactive mode] Up/Down arrow behavior on wrapped rows of a long input line is undocumented
- [DOCS] Managed MCP policy (`allowedMcpServers`/`deniedMcpServers`) enforcement windows are not documented
- [DOCS] Remote Control troubleshooting missing "stuck on reconnecting" entry for OAuth token refresh race during session resume
- [DOCS] Windows startup "Connect to GitHub" popup from Git Credential Manager is undocumented for background git commands
- [DOCS] Statusline docs omit footer-hint (e.g. "esc to interrupt") interaction with a custom status line, and voice-dictation page describes the pre-v2.1.169 behavior
- [DOCS] `claude agents --json` docs omit `--all` flag, new `id` and `state` fields, and which session states the JSON output includes
- [DOCS] Agent view attach/detach docs omit the WSL + Windows Terminal stale/garbled frame that was fixed in v2.1.169
- [DOCS] Background agent docs do not mention the pre-warmed worker dispatch path or its `env` settings behavior
- [DOCS] Plugins reference does not document MCPB plugin cache invalidation and re-extraction behavior on Windows
- [DOCS] [Plugins] `.in_use` PID lock files and the daily stale-marker sweep are not documented in the plugin reference
- [DOCS] OTEL client-certificate env vars docs omit trust confirmation requirement for project settings
- [DOCS] [Dynamic workflows] `/workflows` command docs don't state it opens immediately even while a turn is in progress
- [DOCS] `TaskCreate` tool reference omits v2.1.169 input-repair and unloaded-tool schema-in-error changes
- [DOCS] Errors reference omits the "Your organization has disabled API key authentication" message and v2.1.169 guidance about which credential source to check
- [DOCS] `API_FORCE_IDLE_TIMEOUT` opt-out for the Vertex/Foundry 5-minute idle timeout is undocumented
- [DOCS] [Server-managed settings] Document partial-apply behavior when a delivered payload contains an invalid entry
- [DOCS] `claude -p` slow / appearing to hang on Windows during the slash-command and skill scan is not mentioned in troubleshooting or the headless guide (regression v2.1.161–v2.1.168, fixed in v2.1.169)
- [BUG] Large tool outputs containing Japanese text become corrupted and expand into fabricated lines on Windows
- [Bug] Empty bug report submission
- [BUG] claude agents: cursor not visible in typing window on first session attach (Windows Terminal, Windows 11, 2.1.169)
- Support custom file-opening actions in keybindings (e.g., open settings.json)
- Tool calls intermittently fail with "malformed and could not be parsed"; markup rendered as chat text
- OTLP telemetry silently not emitted on v2.1.153 and v2.1.169 (macOS, interactive TUI)
- [BUG] /model and /effort mutate global settings.json — breaks agents/fleet view; no way to configure per-agent model or effort
- VSCode extension /model picker shows Opus 4.6 as Default instead of Opus 4.8 1M for Claude Max account
- Model acknowledges pattern error explicitly then immediately reverts to same pattern
- [FEATURE] Folders/Bins for normal chat
- Context lost after clicking software upgrade
- [BUG] Windows app update v1.11187.4 silently changed Cowork space model from Sonnet 4.6 to Opus 4.7, model selector now greyed out
- [MODEL] Confabulation: assistant fabricated entire conversation turns and falsely reported ~30 file operations as successful (claude-opus-4-8)
- [Bug] Anthropic API Error: Context Loss Between Tool Calls
- Desktop app shows/selects non-1M Opus 4.8 for a session the CLI reports as 1M
- [BUG]
- I don't have enough information to generate a meaningful GitHub issue title from this message. This appears to be a status update rather than a bug report or feature request. To create an appropriate issue, I would need: - What problem were you experienc
- [BUG]
- [Bug] Image decode fails on GIF: "Failed to decode image: The image format Gif is not supported"
- Feature request: a producer-initiated "resume session" signal — a clean one-shot wake for background/external events
- channels: mid-turn inbound injected into TUI input box instead of queuing as discrete turn
- [JetBrains] Regression: clicking file highlight in chat opens system editor instead of IDE (broke ~v2.1.160)
- Workflow model override: 'use Opus to plan' leaks to all subagents instead of planning phase only
- [Bug] /ultraplan fails on newly-created repos due to CLI repo-sync lag
- [Bug] /copy command corrupts UTF-8 text to latin-1 mojibake in clipboard write
- Feature request: copy button for individual chat messages
- [BUG] Rewind broken in Claude Desktop 1.11187.4 — "rewind target not on CLI's active chain"
- AskUserQuestion dialog covers the content above it — add a collapse/minimize toggle
- alternateScreen: false setting is ignored (v2.1.169)
- [BUG billing] Got silently upgraded to Max plan from Pro without my consent. I got charged 200$ and no refund twice.
- [Bug] @ context file syntax not working in Claude Code agent
- Sonnet 4.6
- [FEATURE] Add intent comments inside generated scripts before execution
- [BUG] Cursor renders in wrong location (outside input box) after submit / new response / resize — vim mode, Linux + macOS
- Plugin-bundled MCP server is double-registered in Cowork; interactive MCP-UI ("app") tools only render via the connector-UUID registration, not the mcp__plugin_… one
- Feature Request: Streaming text output in CLI (typewriter effect like Claude Desktop)
- [FEATURE] [Enhancement] Allow @claude bot to accept repository collaborator invitations
- Desktop app leaks pseudo-terminals (ptys) until the system runs out (forkpty: Device not configured)
- [BUG] Claude's diff editor in VSCode silently discarded my edits to the file
- Regression: background shells killed ~5s after result breaks long-running background commands
- [BUG] enter or shift + enter both submit the query
- Opus 4.7 in Claude Code kicked paid EAS Build without authorization after explicit user rule
- [Bug Report] Unable to process request without additional context
- [BUG] Syntax Highlighting of C# disappears after a brief moment
- Mouse cursor invisible in session after returning from agents view (v2.1.169 regression)
- [Bug] Worktree operations default to main branch instead of worktree branch
- Per-agent frontmatter to skip CLAUDE.md / memory inheritance (the engine already does this for Explore/Plan)
- [BUG] Sonnet usage error on CC
- [FEATURE] show user reject mesage to an edit front and center
- Can hooks observe skill invocations and access the skill name?
- Bug: Stored memory directives (MEMORY.md) are ignored during fast state processing, causing false infrastructure alerts
- [Feature] Allow configuring /goal evaluator timeout and show timeout errors clearly
- [Bug] False-positive cyber-related safeguards block on legitimate low-level storage debugging (SPDK / NVMe-oF vfio-user, read-only gdb)
- [FEATURE]
- [Bug] /copy command missing interactive block picker for code selection
- [BUG] /cost command causes session to freeze on v2.1.158
- [Feature Request] Unified Plan pane aggregating Plan Mode, TaskList, spec documents, and chat plans
- Cross-session message duplication: same user message (identical UUID+ms timestamp) delivered to multiple parallel sessions; duplicate JSONL log lines
- Custom command statusLine never displayed on Windows (Git Bash) — Claude Code 2.1.156
- [BUG] v2.1.113 regression: OAuth initiation ignores `resource_metadata` from `WWW-Authenticate` header, breaking RFC 9728 discovery
- [BUG] /btw side chat opens as regular message instead of overlay
- [FEATURE] claude code api
- [BUG] Cowork: Code block copy button not working in a new window (macOS)
- [MODEL]
- [BUG] Multi-window: "/" skill picker is mispositioned and Enter doesn't select (macOS)
- [Bug] SSH connections prompt for login and fail on keychain-based authentication
- Claude randomly typed some chinese/japanese chars? "Let me do a final状态 check on the tree and tests."
- [BUG] Spell checker uses US English despite system language set to English (United Kingdom)
- [BUG] Gift code "no longer active" after failed redemption while previous gift subscription was still active
- Write tool fails immediately in local VS Code despite Write(**) in permissions.allow — Edit works fine
- [BUG] Inline math ($...$) not rendered in Claude Code interface while display math ($$...$$) works
- [BUG] undefined is not an object (evaluating 'T.data.length')
- Plugin Submission: claude-idle-alert
- [Bug] Claude Cowork fails to start on macOS arm64: Missing Claude Code binary and SSL protocol error
- [BUG] Gamma connector OAuth completes but connector never registers as connected
- Issue with wsl extension
- [BUG]
- VSCode mcp.json gallery and settings.json mcpServers silently launch duplicate MCP server processes for the same name
- Feature Request: autoCompactThreshold setting — auto-trigger /compact before context window is full
- [BUG] /clear drops ultracode (model reverts to non-ultracode) while the corner effort indicator still shows ultracode
- [FEATURE] Usage Bar in Claude Desktop Header + Design Opportunity
- [BUG] Agent still misspells .claude -> .claire creating stray dir on v2.1.169 (recurrence of #31493 / #31984)
- [Feature Request] Option to disable automatic context compression
- Content filtering incorrectly blocking requests involving the Declaration of Independence
- [BUG] Replacement character appears in output
- [BUG] The Notification hook never fires when Claude Code runs as the VSCode native extension
- [Windows] Daemon supervisor runs but never creates its control named pipe — bg service reads as 'not running'
- My legendary shiny cat "fufu" was kidnapped during upgrade. I demand answers.
- [FEATURE] 桌面版输入框布局切换：支持经典模式（输入框跟随对话流）
- [BUG] 2.1.169: interactive sessions write no JSONL transcript (only ai-title stub) — reproduced on WSL2 + macOS; breaks --resume/history
- [BUG] Claude Code VS Code Extension — 400 error when extended thinking IDs are passed back to API
- [BUG] Tool search has broken ranking, causing Claude to fail to find a tool despite exact name match
- [Bug] Unable to process request - insufficient information provided
- Plan mode: add a "save and exit" option that doesn't trigger implementation approval
- [BUG] Claude Desktop app: Remote Control never reattaches existing sessions after a WiFi drop — mints a new auth token per reconnect (99 accumulated, unrevocable); CLI works, Desktop app doesn't
- [BUG] Add --bg as a valid flag to --help to stop Claude Code getting confused all the time
- [BUG] [BUG] [Billing] $2,196 in runaway Auto-recharge charges on Max 5x plan — chat-only user, no Claude Code, no API keys — invoices missing from both claude.ai and Console billing
- [BUG] Model setting in settings.json ignored on new session startup
- [BUG] Desktop — Progressive Renderer Degradation on Windows 11 / NVIDIA G-Sync
- skillOverrides: "off" does not block explicit slash command invocation
- [Windows] Relaunch to Update fails with "Another program is using this file" (MSIX file lock issue)
- Git hook output (pre-commit tests) not surfaced to user — model sees it, user doesn't
- VS Code sidebar loses all session history after restart — sessions-index.json not updated for new sidebar sessions
- No setting to disable inline input autocomplete suggestions (promptSuggestionEnabled does not control this)
- [BUG] Prompt suggestion pre-filled in input box — Enter key does not submit (VS Code extension, WSL)
- Model quality degradation: overconfident false claims, failure to acknowledge corrections, rationalization of incorrect conclusions
- EnterWorktree validates hook-created (non-git) worktrees against the git worktree list, refusing to enter them
- [FEATURE] Session URL appended to commit messages and PR descriptions by default — should be opt-in
- [BUG] PR indicators incorrectly appearing on all sessions in same repository
- System prompt hardcodes 'Claude Opus 4.7 (1M context)' in commit attribution regardless of active model
- [BUG] [Claude Code for VS Code] ESC during IME composition cancels the running task in VS Code extension (root cause located)
- [BUG] Atlassian connector OAuth callback fails - "연결할 수 없음"
- [BUG] Durable CronCreate jobs are project-cwd-scoped and fire into unrelated sessions sharing that directory; tool defaults to durable without confirming scope
- Auto mode unavailable on Pro plan despite valid claude.ai OAuth session
- [FEATURE] Allow configurable OAuth redirect URI (host + path), not just callbackPort, in MCP server definitions
- [BUG] Cowork feature executed PowerShell that removed AMD display driver without user consent
- [Bug] Tool invocation parsing failed: Invalid XML structure in Bash tool call
- [BUG] Report: Claude Code VS Code Extension crashes with exit code 3221225477 on Windows 11 x64
- [BUG] it's a pice of shit!!!
- [BUG] Claude Desktop macOS window stays always on top of other apps
- [BUG] Create button remains grayed out when creating a Briefing Routine with Google Calendar, Gmail, and Slack connectors
- [BUG] Subagents return only "Done" with no output when dispatched for search/research tasks (Opus 4.8)
- [BUG] Keep omn getting the message: Tool result is too large. Maximum size is 1MB.
- [Bug] TDD writes KNOWN_RED tests masking actual CLI exit code failures
- [FEATURE] Register as VS Code inline chat edit tool provider
- [MODEL] Multiple behavioral failures in single session: instruction violations, blame-shifting, in-session memory loss, and unethical problem-hiding (claude-sonnet-4-6)
- [BUG]
- Channel mode: injected draft never auto-submits into a turn, causing watchdog false-restart (~every 30 min, context lost)
- `.claude/` edits prompt on every edit in acceptEdits mode; `permissions.allow` rules and the per-session option don't suppress it (2.1.169)
- [Feature Request] Granular per-category notification preferences (desktop/mobile) + don't suppress blocking prompts on mobile when terminal is focused
- [Bug] Claude Code stops abruptly during code generation without completing changes
- Feature request: concurrent personal + work account sessions
- Missing conversations in desktop version
- [BUG] Claude desktop Windows ARM Version 1.11187.4 (584005) Broke PDF Reading
- I need more information to generate a GitHub issue title. The word "court" alone doesn't describe a bug or feature request related to Claude Code. Could you please provide: - What problem or feature you're reporting? - Any error messages you encountered?
- [BUG] Voice mode unable to access project memory, instructions, and attachments on Claude Web
- [FEATURE] Dockable side chat: embed it as a resizable right-hand panel in the desktop app instead of a floating window
- [BUG] Trackpad scroll on macOS (VS Code terminal): ~3 lines/notch, no scrollbar, distance scales with gesture velocity; CLAUDE_CODE_SCROLL_SPEED=1 no help
- [BUG] Claude Cowork sandbox never becomes ready on Windows ARM64 / Snapdragon X Plus
- [Feature Request] Session limit accounting for skill creation should be separate from general usage limits
- [BUG] Claude Desktop (macOS) ignores MCP tools/list pagination — tools past page 1 never load
- [Bug] TUI rendering corruption in Tmux: overlapping text output and screen state mismatch
- [MODEL] Severe multi-symptom degradation since 2026-06-08 on Opus 4.8 (Desktop app): ignores CLAUDE.md, bypasses permission prompts, hallucinates, refuses doable tasks, writes files unprompted
- [Bug] Windows: terminal/cmd windows flash on every subprocess spawn across multiple CCD sessions
- Smart auto-compact — skip if session is ending
- [BUG] Preview pane auto-refresh steals keyboard focus from chat input while typing (Windows desktop)
- MCP SSE server: tools/call fails with -32602 despite valid connection and correct schemas
- claude.ai web slows down as stale scheduled-routine run cards accumulate (showing 137 'running' tasks)
- ad[Bug] Multi-line code block copy-paste formatting inconsistencies
- [Bug] GitHub MCP Connection Failed
- [BUG] Repeated <ip_reminder> injections in CCR web session instruct model to conceal content from user
- [Bug] Anthropic API Error: Tool result without corresponding tool use block
- Built-in command-safety analyzer false-positives on quoted Windows paths with spaces
- [Bug] Agent view displays incorrect timestamp and status after session deletion
- Context limit clear doesnt do anything
- [Bug] Anthropic API Error: False Positive AUP Violation on Legitimate Academic Manuscript Analysis
- WorktreeCreate hook fails with EPERM for repos under ~/Documents — hook spawned via 'disclaimer' shim loses the app's TCC Documents grant
- Plugin skill names with non-ASCII (e.g. Japanese) characters are sanitized to hyphens, causing identifier collisions and "Unknown skill" on Skill-tool invocation
- [BUG] Claude Code in the Claude Desktop app runs UserPromptSubmit hooks and enforces blocks, but renders no user-facing hook feedback (CLI renders all)
- Remote-control
- Stop hook: $CLAUDE_PLUGIN_ROOT not injected when running plugin hooks
- [High Priority][Billing/Auth] Paid Claude invoice shows Paid but account remains on Free plan
- [BUG] Claude refuses to write CLAUDE.md when it's a symlink
- [BUG] Empty chat, single 'hi' message consumes 47.7k tokens
- Add SessionRename (PreSessionRename / PostSessionRename) hook event
- Feature request: organize conversation/session history into folders or tags
- [BUG] Read tool falsely reports unencrypted pandoc/LaTeX PDFs as 'password-protected'
- [FEATURE] Stop/SubagentStop hook payloads should expose token usage, cost, and duration natively
- [FEATURE]
- Config dialog scrolls to bottom when switching pages instead of maintaining focus position
- [BUG] Claude Code on the web: `us.sentry.io` blocked under Trusted egress (403 `host_not_allowed`) — `*.sentry.io` missing from the enforced allowlist despite being a documented Trusted default
- [Bug] Claude Code ignores explicit implementation instructions and user directives
- [BUG]
- Bash tool (fish): block constructs (`for…end`, `if…end`) fail inline with `parse error near 'end'` though valid in fish
- [BUG] Claude Managed Agent internal proxy throwing 503 when interacting with Git
- [BUG] WSL remote on Windows: false `fork/exec ... no such file or directory` on ccd-cli (binary is present); old chat windows can't reconnect after remote-server restart
- [BUG] Backspace deletes entire word instead of single character in chat input on Windows
- VS Code extension: two-column layout (or height caps) so input + long messages don't cover Claude's replies
- CLAUDE_CODE_PLUGIN_KEEP_MARKETPLACE_ON_FAILURE still breaks fresh marketplace install on 2.1.169 (ref #43929)
- [BUG] Failed payment still upgrades account to Pro plan via Claude Code /login
- [BUG] Top-level custom agents (via --agent / agent setting) do not receive MCP tools; tools: frontmatter ignored
- Voice input (Linux/PulseAudio): audio streamed to STT is ~6× undersampled vs declared 16 kHz → always "no speech detected"
- [Bug] Anthropic API Error: Tool result block missing corresponding tool use block
- [BUG]
- [Bug] Anthropic API Error: Fable Insufficient Permissions for Security Research Use Case
- Terminal UI rendering breaks on Bengali (Bangla) / complex-script text — corrupted gutter, misaligned columns, overlapping glyphs
- [Bug] Fable 5 model safety filters causing false positives on benign messages
- [Bug] Unexpected model switch to Claude Opus 4.8 during task execution
- [Feature Request] Allow biology content for legitimate medical/research use cases
- I don't have any bug report content to generate a GitHub issue title from. Please provide the bug report or issue description, and I'll create an appropriate technical title for you.
- [BUG] No repos match in repo picker despite GitHub app installed and connector connected
- False positive: Usage Policy block on lattice gauge theory question Fable 5, v2.1.170) [BUG] Request ID: req_011Cbt6Zpk3Rup3RBX4CjmLq
- [BUG] Vertical screen tearing/corruption
- [BUG] API Error 400: data retention mode 'default' is not available for this model via AWS Bedrock
- [BUG] Fable 5 safety classifier auto-switches to Opus mid-session during authorized, in-scope security testing
- all sessions show in the sidebar but their message history is empty
- [BUG] /effort auto setting is lost when switching models — silently reset to a fixed level (Fable 5 → low)
- [Bug] Safety filter incorrectly triggers on legitimate cybersecurity development code
- [Bug] Anthropic API Error: Overly Restrictive Security Filtering on Financial Code Analysis
- [Feature Request] Add support for ancient DNA alignment workflows
- [BUG] Scroll wheel intercepted by prompt history navigation — no way to disable
- [BUG] Built-in Co-Authored-By: Claude commit trailer asserts AI co-authorship by default (contrary to U.S. Copyright Office guidance)
- [Feature Request] Actor-based output highlighting: distinguish "user must act" from "assistant is handling"
- Long-session time/state drift: assistant produces wrap-up artifacts with stale 'today/tomorrow' phrasing after date-rollover system-reminder
- Fable 5 Flagged work for cybersecurity or biological topic
- [Bug] Content policy filter incorrectly flags authentication audit request as security-related
- [BUG] Permission allowlist patterns cannot match Windows paths with backslashes
- [BUG] trouble connecting with my remote claude.ai account
- [Bug] Repeated Security Flag Despite Multiple Verification Completions
- Mac terminal: text selection for copying intermittently stops working
- [MODEL] FABLE usage policy
- [BUG] ⏺ API Error: 400 tools.11.model: 'claude-opus-4-8' cannot be used as an advisor when the request model is 'claude-fable-5'.
- [BUG] .worktreeinclude silently skips files with non-ASCII characters in their name
- [Bug] Anthropic API Error: Random "Violate Usage Policy" on Legitimate Code Requests
- [Bug] Vulnerability scanning not working with current model
- [DOCS] [Troubleshooting] Sessions launched from VS Code integrated terminal or any shell that inherited Claude Code env vars — v2.1.170 transcript/resume fix is not mentioned in the troubleshooting or VS Code pages
- Feature suggestion: Culturally diverse / multilingual loading messages in the processing spinner
- [BUG] First prompt lost in new project directory — trust prompt consumes the input, user must type it twice
- [Bug] Claude Code crashes on dynamic workflows causing excessive API token consumption
- [Workflow tool] argsPath: let a file-based / visually-authored spec be the canonical workflow input (SOLO FDE use case)
- [Bug] Incorrect Security Classification of Reversible Hardware Modifications
- "Jump to bottom" indicator obscures conversation text — move it to a corner / make it dismissable
- [Feature Request] Enable Anthropic API security scanning for PHI/healthcare data protection
- [BUG] Account restriction appeal flow redirects to /new after Max upgrade + CVP application
- [Bug Report] Unable to generate issue title - no bug report content provided
- [FEATURE] Claude Desktop Transcript View settings to expand Code changes only
- [FEATURE] Slack plugin: add slack_pin_message / slack_unpin_message tools
- [FEATURE] Slack plugin: add slack_trigger_workflow tool
- [Bug] Norwegian number formatting not following writing conventions
- [FEATURE] Slack plugin: add slack_pin_message / slack_unpin_message tools
- [FEATURE] Slack plugin: add slack_trigger_workflow tool
- [BUG] Claude Desktop installation fails with HRESULT 0x80073CF6 — CoworkVMService blocks install even when stopped, WindowsApps folders undeletable
- [Feature Request] Add security scanning exemptions or allow suppression of false positives
- UI Localization / i18n support — Portuguese (pt-BR) and other languages
- [Bug] Anthropic API Error: Usage Policy violation on reverse engineering detection
- [Bug] Anthropic API: Overly broad cybersecurity classifier flags defensive vulnerability redaction as offensive content
- [BUG] Fable 5 safety classifier auto-switches to Opus mid-session during authorized, in-scope security testing
- [FEATURE] persist session color (/color) across sessions via settings.json
- [BUG] claude-api skill trigger self-fires on /model banner (bare model names + ANSI [1m as trigger tokens)
- [Bug] Anthropic API Error: Server rate limiting requests
- YURI production Flagged for some reason while using fable
- [BUG] Subagent dispatch prompt and returned result hidden in main thread — collapsed line cannot be expanded (regression after 2.1.160, still in 2.1.170)
- [BUG] scheduled-tasks MCP cron-spawned claude.exe never exits on Windows — accumulates leaked processes (~350 MB each)
- [BUG] Regression (v2.1.160+): subagent spawning fails on Bedrock with "thinking.type must be enabled or adaptive" for Claude 4.x models
- UserPromptSubmit hook timeout is hardcoded to 30s (HN4=30000) with no override; context silently dropped on timeout
- [Bug] Router unexpectedly switched model from Claude Fable 5 to Opus 4.8
- [Bug] CVP-enrolled repo flagged for security review despite receive-only RF tool compliance
- [Bug] False positive cyber violation detection blocking legitimate code review
- [Bug] Inappropriate content filter blocking legitimate research review requests
- Fable 5 refuses to assist with 'Application Security Architect resume' editing
- [Bug] Mythos flag incorrectly triggered during connection method selection
- Fable 5 safety classifier fires model_refusal_fallback on a bare `hello` — scores the static request preamble, not the user message
- [MODEL] Your Fable 5 is officially a waste of fucking water, congratulations!
- [Feature Request] Prevent model from attempting to decompile Claude Code binary
- Bash permission patterns do not support ** glob — impossible to scope by both path depth and file extension
- [FEATURE] route specific sub-agents to a different API endpoint/key
- [BUG] [MODEL] False positives: safety system blocks benign system-administration / self-hosting tasks
- [BUG] Mobile remote-control model picker shows stale selection after /model change in connected CLI session
- [Bug] Model unexpectedly switched to Opus during security audit task
- [Bug] Anthropic API Error: Content Policy Violation on Security Review Request
- --remote-control: model selection in Claude app does not propagate to remote session
- [BUG] claude code TUI has gone down hill
- [Bug] Fable 5 security classifier incorrectly flags legitimate CSRF/HMAC-protected code as threat
- [Bug] Unable to process empty bug report
- [Bug] Model unexpectedly switches to opus-4.8 when using fable-5
- [Bug] Fable 5 model safety measures blocking normal conversation content
- [Bug] Anthropic API Error: Model triggered security review causing request rejection
- [Feature Request] Allow Claude Code to integrate with security hardening tools like Fable
- [Bug] Fable 5 bio classifier blocks innocuous academic statistics terms (组学, spatial transcriptomics)
- [Bug Report] Unable to process request - no details provided
- [BUG]
- [Feature Request] Display task summary and specifications before long-running operations
- I need more information to generate a GitHub issue title. Please provide details about the bug or problem you're experiencing, such as: - What were you trying to do? - What error message did you receive (if any)? - What happened vs. what you expected? - A
- [BUG] Desktop: computer-use side-panel dock fires phantom quit event; "Wait for Claude" silently quits app after task; no setting to disable docking (multi-monitor)
- [Bug] Fable 5 model incorrectly flagged as unsafe, forcing unnecessary Opus 4.8 fallback
- [Bug] Missing Fable 5 option in /advisor command choices
- [Bug] Advisor service unavailable
- [BUG] Bun startup segfault on Windows 11 26200 + Intel Meteor Lake (Core Ultra 5 135U) — reproduced on current 2.1.170; prior report #55219 auto-closed as stale
- No matter what I type it says"violate our Usage Policy"
- [Bug] Claude Fable 5 classifier false positive on standard refactoring commands with security-unrelated terminology
- [FEATURE] Let Workflow (dynamic workflow) subagents render in tmux panes, like teammateMode: "tmux" does for agent teams
- [Feature Request] Enable penetration testing of user code locally
- [Bug] Unable to process empty bug report
- [BUG] "Thinking" text bullet is half-white when the text is a shade of yellow
- [Bug] Anthropic API Error: Overly Aggressive Safety Filtering on Non-Malicious Content
- [Bug] Mythos incorrectly flagging non-cybersecurity content as security-related
- [BUG] When user prompted with suggested prompt, pushing return key fails to submit
- [Bug] Security issue detection fails after user confirmation to fix
- [BUG] UI shows "update available" for latest-channel versions when installed from stable apt repo
- [Bug] Unable to respond to clarifying questions in multi-turn security audit conversations
- [BUG] Every single attempt at reviewing my codebase triggers fall back to opus 4.8
- [Bug] Fable 5 safety classifier false-positives on authorized defensive security audits
- [Bug] Workflow tool accepts arguments without validation, causing silent failures with wasted API calls
- [Bug] Anthropic API Error: Safety Flag Blocking Legitimate CTF Usage
- Model fabricates <task-notification> blocks (plus its own replies) inside a single assistant message during long background-task waits
- [Bug] Anthropic API Error: Overly Restrictive Content Policy on Abstract Mathematical Simulations
- [Bug] Skill-specific memory not persisting across session restarts
- [Feature Request] Dynamic workflows should auto-select appropriate model instead of inheriting parent model
- Stop hook does not fire when awaiting AskUserQuestion response
- [BUG] "--remote" web sessions never connect to GitHub remote repo
- Feature request: Codex-style background computer control with concurrent window/session ownership
- [Feature Request] Allow custom naming for git worktrees instead of random generation
- [BUG] Cyber-security usage detected during normal work (non-security related).
- [Bug] Unable to process empty bug report
- [Bug] Claude Code refuses to engage with legitimate security researcher queries
- [MODEL] Opus 4.8: runaway extended thinking (20k-64k output tokens/turn), replies to hallucinated user messages, fabricates forensic "evidence" when asked to investigate
- [BUG] Welcome banner shows subscription default model instead of settings.json-pinned model
- [Feature Request] Whitelist repository for standard cellpose/cupy 3D cell segmentation workflow
- Advisor/model mismatch (400) silently hidden by fallbackModel — misleading "could not be retried" on Fable 5
- Withdrawn
- [BUG] Slash command autocomplete stops filtering after a '.' (period) in skill/command name
- [Bug] False Positive Rate in Content Classifier for Data Pipeline
- [Bug] Biology Safety Filter triggering prematurely on health data analysis
- [Bug] Security scanner falsely flags user code during development
- [Feature Request] Allow Claude Code to assist with mold detection model development and validation
- [Bug] Anthropic API Error: Tool result missing due to internal error
- [BUG] Output content filter blocks legitimate code-signing build scripts (false positive on Authenticode / Microsoft Store signing)
- [Bug] Automatic model fallback to Opus without user consent during execution
Resolved issues (194)
- Hosted Notion MCP (DCR): client re-registers per session → repeated full re-auth (orphaned refresh tokens)
- [BUG] /compact 95% forever
- Per-agent frontmatter to skip CLAUDE.md / memory inheritance (the engine already does this for Explore/Plan)
- Issue with wsl extension
- [BUG] MCP servers from marketplace no longer available
- [Feature Request] Support MCP servers with Client ID Metadata Documents (CIMD) OAuth authentication
- [BUG] Agent-teams teammate panes wedged with "<Box> can't be nested inside <Text>" Ink error on 2.1.119 — not reproducible on 2.1.138 (regression record + fix confirmation)
- [BUG] Stdio MCP tool calls default to 1-second timeout in v2.1.148 — long-running tools fail before server can respond
- [BUG] Rider MCP Server fails to detect installed Claude Code, but does detect Codex, which I've never installed or used
- [BUG] claude.ai MT Newswires MCP connector — fetch returns HTTPStatusError.__init__() crash on all data requests
- Slack MCP via browser session tokens (xoxc/xoxd) unreliable on Enterprise Grid — request first-class OAuth-based Slack MCP
- [Atlassian MCP] Add getJiraIssueWatchers tool
- Zoho Projects MCP connector returns HTTP 500 on all API calls for Zoho CRM Plus accounts
- [FEATURE] Plugin: knowledge-graph — Persistent memory layer with zero-interrupt architecture
- [BUG] Persistent MCP server diagnostic alert after removing configuration
- [FEATURE] [VSCODE] Allow fast account toggle and show current logged in email
- [Bug] Agent execution failures without retry mechanisms
- CLAUDE.md instruction for Japanese conversation titles is not respected by the auto-title feature
- [FEATURE] A PAUSE (Ctrl-Z / suspend) for the agent, distinct from STOP (kill)
- [BUG] Claude doesn't use claude.md
- [FEATURE]
- [Bug] Claude Code ignoring user instructions or refusing to execute commands
- You've hit your weekly limit · resets Jun 7, 4pm (Europe/Rome)
- Expose input-box line-editing actions (deleteToLineStart/deleteWord) as bindable keybindings; add default Cmd+Backspace = delete-to-line-start
- Claude presented guesses as factual knowledge - dishonesty in conversation
- [Bug] Claude Code hangs with repeated "court" string output on each turn
- [BUG] '!yes' crashes claude and fills temporary storage
- [FEATURE] VSCode extension: session titles truncated in Ctrl+Tab switcher despite available screen space
- [BUG] VS Code Extension Not Connecting to Claude
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [BUG] Conector Figma deshabilitado en Cowork — flujo OAuth no funciona
- Sycophantic drift under correction causes loss of precision when user is frustrated
- [Bug] Auto-update Failed Error
- [FEATURE] Feature Request: Optional sync between Claude Chat Projects and Claude Code memory/instructions
- Image/screenshot attachments not delivered to the model (Claude Code on claude.ai)
- [BUG] Claude Code process exited with code 3
- [BUG] Model calls AskUserQuestion with no preceding text block — explanation stays in thinking and never reaches the user
- [Bug] Feedback prompts appearing unexpectedly during tool responses
- [Feature Request] Tab autocomplete should complete to longest common prefix, not full-accept first match
- [BUG] Failed to install plugin" i Cowork på macOS
- [Bug] Anthropic API Error: Tool Result Block Missing Corresponding Tool Use Block
- VS Code extension: managing multiple chats is unreliable — no conversation list indicators, no archiving
- Native binary approach breaks Claude Code on QNX and BSD platforms
- Render statusline before MCP/plugin initialization completes
- Tool call serialized as text (court<invoke>) with stop_reason=end_turn — model emits no executable tool_use, turn silently stalls (claude-opus-4-8)
- Model quality feedback: Sonnet 4.6 misinterprets task instructions and gives nonsensical responses
- [FEATURE] Page Up and Page Down buttons don't work in Cowork unlike in Code - Claude Desktop Windows
- [Feature Request] Prevent auto-adding Claude as commit coauthor
- [BUG] Cowork archive uploads (zip etc.) fail to persist to VM — docx and other regular files work fine
- [Desktop] Respect statusLine setting (run statusline.sh) in the desktop app
- [Bug] Claude Code process frozen unresponsive to user input
- [Bug] Chat fails to execute simple git command, incorrectly reported as completed
- Desktop app (Windows): @ file mention has no autocomplete and folder picker cannot navigate folders
- [BUG] Shift+click required for text selection — regression from ~2 weeks ago
- Cowork (macOS): VM SDK 2.1.163 fails download_and_sdk_prepare ("Download failed") though artifact is reachable; no fallback to working 2.1.161
- Allow hiding/removing unused cloud connectors (Gmail, Google Calendar, etc.)
- [FEATURE] Microsoft 365 MCP:Add support for reading email attachments
- Spawned child process (versioned binary) grew to 12.3 GB RSS and was OOM-killed
- MCP tool-call parser appends `</payload></invoke>` to string parameter values (100% repro within affected session)
- [BUG]
- tools/list_changed cannot recover stale Python subprocess code
- Add setting to disable auto-attaching IDE selection/active file by default (VS Code extension)
- [BUG] Cowork Chrome MCP connection failure - persistent "not connected" error
- [Bug] Model refuses to execute git add command despite enabled tool permissions
- [BUG] Slack tools return empty `Text:` for bot messages using legacy `attachments[]` field
- Usage Policy refusal cascades into OAuth session invalidation — 5 forced re-logins/day during defensive coding on own codebase
- Telegram plugin MCP connection drops after each conversation turn
- [BUG] AddPackage failed: HRESULT 0x80073CF6
- bug: Task subagents receive empty MCP tool surface (main-thread works, subagent context drops all mcp__* tools) — Windows + Opus 4.7
- MCP gateway WAF blocks legitimate tool-call bodies (SQL/shell/config text)
- Windows: 杀毒软件的"PowerShell 脚本执行检测"导致 Claude Code Shell 全部 EPERM
- [BUG]
- [Bug] MCP processes not properly cleaned up on exit, causing zombie processes and memory leak
- [Bug] Unable to process request - insufficient information provided
- [BUG] You messed up scroll back in one of the latest updates
- [DOCS]
- [BUG] Claude for Office (PowerPoint/Excel), Mac Broken
- False commitment in code implementation
- [Bug] Crash related to Bun runtime
- [Bug Report] Unable to process - no issue details provided
- [Bug Report] Unable to generate issue title - no bug report content provided
- [BUG] file_upload missing in claude-in-chrome mcp
- Bootstrapped Agentic OS as Claude Code Default Onboarding
- Feature request: flag to render Skill invocations compactly when the skill is already loaded
- [BUG] Weekly Sonnet limit non-functional — blocked despite 78% Sonnet quota remaining
- [BUG] Extra usage fully consumed despite dashboard showing 43% remaining — Claude Code for VS Code
- [BUG] Long responses show repeated text chunks in scrollback on Windows 11 (PowerShell)
- [BUG] `memory:` field in subagent frontmatter not functional — reproduces on v2.1.137; tools allowlist appears to override auto-enable
- [BUG] PreToolUse hook "ask" permissionDecision is auto-approved in auto mode — no way to force prompt
- Question: trailing `&` to dispatch a backgrounded prompt — is this a supported feature?
- [BUG]
- [BUG] SendMessage silently fails when using agent name — only agent ID works for resuming completed agents
- Auto-update silently installs truncated native binary on darwin-arm64 when laptop sleeps mid-download (v2.1.138)
- Status bar shows wrong effort level after /effort max
- [BUG] "PDF too large" error incorrectly triggered on plain text (.txt) files in VS Code extension
- Feature: Landlock sandbox directives for PreToolUse hooks
- [Feature Request] Add auto-complete functionality to Claude Code CLI
- [BUG] Cowork "New Project" dialog shows mixed path separators on Windows
- [BUG] Windows Desktop app + SSO (Team plan): oauthAccount not hydrated to ~/.claude.json, OTel metrics emitted without user identity
- [Bug] Web client OTLP telemetry missing user.account_id, user.email, and organization.id fields
- [BUG] claude -p headless: no spend visibility/cap for orchestrated workloads -- $313 burned in 8.5h on a single retry-stuck item; support routing loop with 3 Fin contradictions
- [Opus] Claude Code reported nohup process as "aborted" but never killed it per my instructions--$350 in unauthorized API charges over 5 days
- Scheduled tasks stop firing autonomously after 12-24 hours; manual reboot triggers catch-up burst that swallows queued runs
- [BUG] Cowork scheduler stops firing recurring tasks after ~24-30 hrs uptime; only Windows reboot recovers
- [BUG] Model setting in settings.json ignored on new session startup
- [FEATURE] VS Code extension edits bypass the editor API, losing Timeline/Local History entries
- Claude Desktop - Unknown Error on startup after computer restart (Windows)
- [BUG] PR indicators incorrectly appearing on all sessions in same repository
- Channel mode: injected draft never auto-submits into a turn, causing watchdog false-restart (~every 30 min, context lost)
- Telegram plugin (0.0.6) inbound inject silently dropped — CC host doesn't declare experimental.claude/channel capability
- [BUG] Claude Desktop doesn't connect to Custom MCPs altogether (not with OAuth 2.1 nor with SSE)
- [BUG] Keep omn getting the message: Tool result is too large. Maximum size is 1MB.
- [BUG] cant remove old Routine session runs which enforce Routine limit which is not valid
- [BUG] /resume command fails with "Invalid `signature` in `thinking` block" error when previous session used Extended Thinking
- [BUG] Top-level custom agents (via --agent / agent setting) do not receive MCP tools; tools: frontmatter ignored
- [Bug] Anthropic API Error: Fable Insufficient Permissions for Security Research Use Case
- [BUG] /effort auto setting is lost when switching models — silently reset to a fixed level (Fable 5 → low)
- [BUG] Permission allowlist patterns cannot match Windows paths with backslashes
- [Bug] Repeated Security Flag Despite Multiple Verification Completions
- [Bug] Content policy filter incorrectly flags authentication audit request as security-related
- Fable 5 Flagged work for cybersecurity or biological topic
- [Bug] Safety filter incorrectly triggers on legitimate cybersecurity development code
- [Bug] Anthropic API Error: Overly Restrictive Security Filtering on Financial Code Analysis
- [Bug] Anthropic API Error: Random "Violate Usage Policy" on Legitimate Code Requests
- [BUG] First prompt lost in new project directory — trust prompt consumes the input, user must type it twice
- I don't have any bug report content to generate a GitHub issue title from. Please provide the bug report or issue description, and I'll create an appropriate technical title for you.
- [FEATURE] Slack plugin: add slack_pin_message / slack_unpin_message tools
- [FEATURE] Slack plugin: add slack_trigger_workflow tool
- [Bug] Usage Policy classifier stuck in blocked state for legitimate kernel security work
- [BUG] Claude Desktop installation fails with HRESULT 0x80073CF6 — CoworkVMService blocks install even when stopped, WindowsApps folders undeletable
- [BUG] 2.1.169: interactive sessions write no JSONL transcript (only ai-title stub) — reproduced on WSL2 + macOS; breaks --resume/history
- [Bug] Router unexpectedly switched model from Claude Fable 5 to Opus 4.8
- [BUG] scheduled-tasks MCP cron-spawned claude.exe never exits on Windows — accumulates leaked processes (~350 MB each)
- [BUG] Mobile remote-control model picker shows stale selection after /model change in connected CLI session
- `/loop` skill loads in web sessions but its backing tools (`CronCreate` / `ScheduleWakeup`) are not loadable
- [BUG] Claude for PowerPoint Add-in: "input_schema does not support oneOf, allOf, or anyOf at the top level" — completely unusable
- [BUG] tarda de 10-20 minutos para resolver un problema!
- [Bug] Claude Code not respecting user-defined rules and constraints inconsistently
- [BUG] Claude Code with Auto Mode abandone skill, act on its own
- [Bug] Excessive verbosity in unprompted detailed responses
- [Bug] Simultaneous feedback request and number input prompt causes UX conflict
- [Bug Report] Unable to process issue - no bug report provided
- I'm ready to help generate a GitHub issue title for Claude Code. However, I don't see a bug report in your message - you've provided the instructions but no actual bug report to analyze. Please share the bug report details, and I'll generate an appropriat
- [BUG] API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
- [Bug] Claude ignoring instructions from CLAUDE.md configuration file
- [BUG] claude-agent-acp: streaming loop hangs on missing session_state_changed:idle (stop button stuck) — UPDATED with proper bounded-drain fix
- [Feature Request] Improve task completion tracking and prevent skipped tasks
- [BUG] Jetbrains Plugin (Rider and Pycharm) Empty File causes Cancel
- Self-spoofed user instructions: assistant emits 'Human:'-prefixed text in long-context background turns, executed as real user input
- Claude loses in-session context at low utilization, forgets recent decisions and instructions
- Project-scope plugin auto-discovery from <cwd>/.claude/plugins/*/
- Desktop app's bundled Claude Code 2.1.111 exits with code 1 immediately on prompt submission (macOS)
- [BUG] claude --resume <id> cannot find intact transcript after original worktree path is removed
- [Bug] Permission prompts not respected when auto-edit or bypass permissions toggle enabled
- Make /insights a first-class workflow primitive (3 asks)
- [BUG] CLAUDE_CODE_ADDITIONAL_DIRECTORIES_CLAUDE_MD=1 does not load .claude/rules/*.md from --add-dir paths
- [BUG] VSCode: hiding a session hides it in every project that shares its sessionId (global hiddenSessionIds, no project scope)
- [Bug] Timezone handling not respecting user location (Denver/MT instead of UTC)
- Feature: opt-in suppression of CLAUDE.md / skill-catalog injection on subagent dispatch
- [BUG] remote control command unavailable on pro account using claude code version v2.1.133
- [BUG]
- [Windows/Git Bash] PreToolUse hook enforcement silently bypassed — hooks exit 49 with no output or warning when python3 resolves to Store stub
- MCP OAuth SDK sends empty User-Agent, causing 403 from WAF/Cloudflare-protected servers
- [BUG] Chrome extension fails to connect on Windows 11: MSIX package ACL blocks native host execution (os error 5)
- [BUG] without confirmation,ran find across entire Desktop and sent personal file to a 3rd-party API
- [BUG] Max plan usage at 100% with zero requests — claude.ai (not Claude Code)
- Add setting to suppress the recurring auto-mode explainer banner
- RFC: Deterministic tool gate — hooks are necessary but insufficient for governance enforcement
- Memory base directory should follow project context when navigating between projects in a session
- [BUG] generate_review.py embeds JSON into <script> tag without escaping < — breaks eval review page when outputs contain HTML/JSX code examples
- [BUG] settings.json `env` block not injected into hook subprocesses when running with --output-format stream-json (e.g. Cursor extension)
- [FEAT] Auto theme should work with colorblind-friendly palettes
- Desktop app: add split panes and numbered tab shortcuts (Cmd-1/2/3) for power users
- [BUG] Why has the auto-compaction mechanism been triggering recently whenever the conversation context reaches around 50–60%?
- 🚪🤜 KNOCK KNOCK...
- [Bug] /slash command fails to detect external file changes
- [BUG] BUG: Cannot add UNC network path as project folder on Windows
- Branch status badge shows diff-vs-base line count instead of branch content
- [BUG] swift-lsp pushes stale 'Cannot find type/module' diagnostics — #17979 fix not extended to SourceKit-LSP
- Remote Control sessions show "Remote Control disconnected" banner on first turn despite session succeeding (control_response error="Already initialized")
- Claude Opus 4.6 repeatedly ignored user instructions, wasted hours of credits
- [BUG] Markdown preview pane doesn't support intra-doc anchor navigation
- [FEATURE] Autopsy feature
- [BUG] Claude Code Dispatch session remained functional but I lost all visibility into it from the Mac desktop UI.
- [BUG] Intermittent Korean syllable corruption in AskUserQuestion — wrong valid characters generated (not U+FFFD)
- Statusline: context_window.used_percentage always 0 on native Windows (Git Bash)
- Feature: Allow whitelisting sites (including localhost) in Claude in Chrome extension
- [FEATURE] Desktop app Code tab: cap conversation column max-width for readability on wide windows
- [Bug] Anthropic API Error: Content filtering blocks benign state dropdown feature request
- [Bug] Cyrillic characters render as question marks due to font encoding issue
- [Bug] Anthropic API Error: Safety Flag Blocking Legitimate CTF Usage
- [DOCS] Tools reference missing Advisor Tool documentation
- [BUG] Preview pane auto-refresh steals keyboard focus from chat input while typing (Windows desktop)
- Withdrawn