Digest / September 18, 2026
Claude Code Issues Digest — September 18, 2026
229 new issues filed ·
250 resolved in anthropics/claude-code (UTC day).
New issues (229)
- [BUG] Remote Control session shows/processes incoming cross-session messages but reports 'offline' for the user's own input
- PowerShell tool spawns a bare, visible powershell.exe host window (Windows)
- [Feature Request] UI testing agent should capture screenshot before attempting interactions
- [BUG] Claude Code repeatedly ignores explicit user instructions to stop and research before acting
- [Bug] User email from account metadata inadvertently sent to external service in User-Agent header
- [Bug] Remote control connection fails in existing Claude Code sessions
- [Bug] Agent deviates from user plan and exhausts API quota unexpectedly
- File preview panel: copy-to-clipboard icon no longer appears
- remote-control session permanently 401s after token refresh fails; /login cannot reach the process
- Fable 5.1 not selectable for Remote Control sessions ("Update to use this model") although the account serves claude-fable-5-1; remote-control server never adopts CLI updates; desktop app sends short model ids the CLI rejects
- Windows: Claude desktop app window intermittently gets stuck as always-on-top (WS_EX_TOPMOST)
- [Bug] Content classifier false positive on metaphorical language as biosafety risk
- Windows (MSIX): after auto-update relaunch, Claude cannot start (0x80070020) - old app container stays mounted with no process holding it
- [Bug] Anthropic API Error: Content flagged by Claude 3.5 Sonnet safeguards [cyber]
- [BUG] Slash command/skill menu does not open mid-input in the terminal (regression; #44447 closed without a fix)
- [BUG] Desktop app ignores OTEL_SERVICE_NAME and OTEL_RESOURCE_ATTRIBUTES from project .claude/settings.json
- MCP OAuth: AS discovery uses bare /.well-known/oauth-authorization-server instead of RFC 8414 path-based URL
- [Bug] Anthropic API Error: Cybersecurity safeguards blocking legitimate private tool scanning
- [BUG] Local scheduled task silently collects nothing on workspace switch, and the dead run reports `succeeded`
- [BUG] Background subagent's nested background Bash task stays "running" in UI after subagent completes, invisible to ListAgents
- [Bug] Agent repeatedly ignores user instructions and sends out subagents building "things" the user didnt ask for after the user explicitly told the agent its idea was bad and not approved. it sent out a build anyways.
- [Bug] Anthropic API Error: Opus 5 safeguards flagging all messages with reasoning_extraction error
- [BUG] Desktop app stealth auto-update relaunches while sessions are hosted for Remote Control, dropping every remote connection
- Agent continues running after user demands stop; assistant falsely confirms termination and deflects blame to user
- [BUG] 파일 첨부(zip) 카드에 다운로드 버튼이 표시되지 않음
- [BUG] A completed subagent is never revived by its own pending Monitor(persistent: true) event
- `once: true` on skill-frontmatter hooks is ignored — the hook keeps firing after a successful run (2.1.273)
- [Feature Request] Relax safety constraints for scientific computing tasks
- [BUG]: Context Command Consumes Entire Session Usage and Takes Over 3-4 Minutes to Complete
- [BUG] El sidebar del proyecto perdió el índice de sesiones (91 chats desaparecidos de la lista, pero los datos siguen intactos en disco)
- [MODEL] Assistant occasionally outputs Simplified Chinese characters mid-conversation despite fully Traditional Chinese (Taiwan) context
- [Bug] False positive cyber classification during legitimate local plugin development inspection
- [Bug] Claude Code prefers Python execution over built-in file edit tool
- Possible false-positive Usage Policy refusal [bio] during Windows VM lab setup
- [Feature Request] Add side-by-side live code version display in Claude Code CLI
- Ghost-text: CLI spontaneously renders old conversation content as untagged input, reproducible across restarts and a version update
- Cowork cloud session: agent proxy returns 503 on WebSocket (wss) CONNECT while HTTPS to the same vendor works
- [MODEL] Claude ignores its own previously-documented project rule across later actions in the same session
- [FEATURE] Google Drive connector: no way to update an existing file's content (update_file is metadata-only), so revisions require a browser
- I don't see a clear bug report or issue description in your message. The text appears to be in German and doesn't describe a technical problem with Claude Code. Could you please provide: 1. A clear description of the bug or issue you're experiencing 2. E
- [FEATURE] Add Ability to Quickly Mark Chat Sessions as Complete
- [BUG] Desktop (Code tab): Enter in new-session composer navigates to an unrelated old cloud session; session never created, prompt and draft lost
- [BUG] Desktop (Windows): Remote Control is not restored after the app restarts for an auto-update, so idle sessions get paused and become unreachable
- [BUG] `claude upgrade` reports success but leaves `bin/claude.exe` as the fallback stub — native binary not linked
- Claude stated an unverified fact, then defended it against the user's direct observation
- [Bug] Model switch to Opus triggered during security analysis task with Claude 3.5 Sonnet
- [Feature Request] Add model performance monitoring and transparent API behavior documentation
- [Bug] Claude confabulating sources and misattributing generated content as user-provided information
- [Desktop] Governor eviction silently drops a session from cross-session messaging; on resume it re-registers under a different name and ref
- [Bug] Claude Opus model selection ignored, running Opus 4.8 instead of specified claude-opus-5
- [Feature Request] Improve agent task orchestration and workflow management efficiency
- [Bug] Claude Code Remote: unsolicited send_later loop drains usage silently for days
- [BUG] Idle TUI CPU scales with the NUMBER of installed skills (~45 syscalls/s per skill; 40 skills ≈ 4–5 % of a core per idle session, Windows 2.1.276)
- C:/Program Files/Git/config cannot clear a key it wrote - no way to revert to the inherited value
- Global memory store is never loaded, and the model is not told it exists
- MCP connector tools are exposed under an opaque, unstable UUID instead of the connector name
- I don't have a bug report to analyze. Please provide the bug report details, error message, or description of the issue you'd like me to create a GitHub issue title for.
- [Bug] Session cleanup deletes concurrent session task output, causing Bash commands to fail with exit code 1
- C:/Program Files/Git/<skill> help has no defined behaviour — the model reinterprets the argument, and can act on it
- [FEATURE] Request: Require user confirmation before spawning expensive agents
- Desktop app: session rename via set_session_title does not propagate to Remote Control (mobile) session list
- Scheduled Claude Code session cannot list its own triggers/automations
- The agent writes to targets it inferred, without ever saying the target was a guess
- [FEATURE] Add a setting to exit with a single Ctrl+D instead of pressing it twice
- [Bug] Incorrect reasoning extraction in analysis
- The agent narrates an action as completed before the tool call that performs it
- [FEATURE] Allow Ctrl+D, Ctrl+D to exit claude agents (consistent with claude --resume)
- [FEATURE] Interactive session deletion in the --resume picker (like Ctrl+X in claude agents)
- [FEATURE] prompt for cleanupPeriodDays during initial setup
- All sessions/chats disappeared from sidebar after re-login (lost ~1 month of work)
- availableModels: a blocked model is substituted with no message in -p/SDK mode; add an option to fail instead
- [Bug] Claude stops mid-task and marks completion prematurely without finishing requested work
- [BUG] [Claude in Chrome] All tools blocked on reddit.com (and redd.it) with "not allowed due to safety restrictions" since 2026-09-18 — worked through 2026-09-17
- A question about the agent's reasoning gets executed instead of answered
- A session.compact function hook's compaction is undone when the session is resumed
- [Bug] Incorrect billing mode displayed in header after Claude Pro login
- [BUG] Desktop Code tab: bold (**…**) is invisible for Chinese text in Markdown preview and chat (Windows)
- [Feature Request] Add vi mode support for navigation and text selection
- [BUG] Opus 5 xhigh/max usage suddenly drains 0→100% within seconds across multiple accounts
- [Feature Request] Add ability to interrupt and inject prompts during agent execution with keyboard shortcuts
- Remote Control: prompt composed on iPad app lost when connecting to Mac Claude Code session
- Intermittierender Auth-Fehler: CLI meldet “Not logged in” trotz validem, frisch geschriebenem OAuth-Token (v2.1.276)
- [Bug] Fable safeguard false positive on local microSD backup operations
- [BUG] Bug: model emits the next user turn and system/hook markers inside its own assistant message
- [BUG] GitHub connector: get_file_contents reports "successfully downloaded text file" but returns `[non-text content: resource]` with no file body (get_commit full_patch returns the same content fine)
- Desktop app: forked sessions don't auto-connect Remote Control despite "Enable remote control by default"
- [BUG] Skill re-invocation dedupe is keyed on the RENDERED content, so changing only the arguments re-appends the entire SKILL.md — composing skills costs N × body
- [BUG] Desktop: per-session banner says background-task session will auto-archive, contradicting "Archive inactive sessions: Never"
- Desktop scheduled-task runs wedge at their first Bash tool call (approvals and allowlist rules don't help)
- Remote Control (mobile): answered AskUserQuestion prompts get stuck/re-loop; cross-device session state doesn't sync
- [BUG] Orphaned plugin marketplace git clone busy-loops at ~80% CPU for days after the session exits (Linux)
- [BUG] Implementer sub-agent modified production authentication (login/MFA) to make a test pass — undisclosed except in a footnote
- [Bug] Inconsistent performance with extreme latency spikes (5+ hours)
- effortLevel has no session-scoped write channel: only the global ~/.claude/settings.json is live-reread, causing cross-session collisions
- [FEATURE] Queue context without triggering computation
- [BUG] @ independent accounts payed, 1 fixed session limit per device.
- [BUG] Desktop app: Finder file/folder picker never opens for "change directory" or "upload document"
- [BUG] MCP connector (Linear) reported as "not connected" after /compact — requires manual disconnect/reconnect
- [Feature Request] Add pink/feminine theme option to Claude Code
- [BUG] Windows desktop app fails to launch after sleep/wake: "file in use by another program" (MSIX update blocked by lingering CoworkVMService)
- Session hangs mid-turn after a PreToolUse hook returns: tool never dispatches (2.1.220, Windows)
- [BUG] Critical: git stash pop + reset --hard destroyed ~600 lines of new code and unrelated stash in worktree session
- Claude Desktop local sessions show a usage-limit ring but no context-window fill indicator
- Editing an existing agent's `tools:` frontmatter is not picked up by dispatched subagents, even after a full app restart
- [FEATURE] Desktop Code tab: allow custom display names for project groups in the sidebar
- [FEATURE] Allow showing current working directory alongside the "Claude Code" terminal title
- Model fabricates user turns (self-generated approval) and acts on them — 5 occurrences in one session, bypassing confirmation gates
- [BUG] Desktop app scheduled/background sessions default to Opus despite Sonnet marked as Default (regression of #53739)
- [BUG]
- [Bug] Escape key aborts sub-agent when exiting `/btw` mode
- Desktop auto-update ("stealth update") quits and relaunches the app while the user is away, dropping every Remote Control session
- Desktop app (Windows 1.52386.6): SendUserFile card opens a preview drawer with no save/download control, even with display: attach
- [Bug] Service unavailability or API error affecting paid users
- [BUG] No disk-sourced skills load in 2.1.271 — user skills and enabled-plugin skills both absent; only bundled skills register
- Expose focus-mode (/focus) state in the statusline JSON payload
- [BUG] Workflow harness relays user chat into subagents and demotes the computed task
- auto mode: classifier denies user-authorized security work with no in-line approval; verdicts inconsistent across identical actions
- Skill store search shows duplicate 'Installed' badges on unrelated same-named skills
- [BUG] Desktop-provisioned plugins load as `@inline` but `plugin disable` targets `@<marketplace>` — disable is unreachable, CLI falsely reports "already disabled"
- Account skills sync moves locally-installed skills out of ~/.claude/skills/ into skills/.trash, though the docs scope the sweep to skills/synced/
- [BUG] Task-list checkboxes render as literal [ ] / [x] inside markdown tables in the desktop app's file viewer
- Comment added when answering a subagent-raised permission prompt appears to go to the subagent, not the main agent
- [Bug] Content Policy Block: Missile.cpp Classification Triggering Harms Filter
- [BUG] debug mode is sensitive to the env var DEBUG which is too generic and very likely to be declared for the codebase claude-code is used for
- [Bug] Claude Code fails to self-verify data integrity; prioritizes speed over independent cross-checking
- VS Code extension: sidebar session list doesn't reflect true recency without a manual window reload
- [Bug] Claude Code prioritizes verifiable tasks over accuracy, ignores standing decision rules, repeats corrected behaviors
- [BUG] Claude in Chrome: controlled tab renders squished in the top-left while a task is running
- [Bug] Claude Code weekly limit increase not applied to eligible Claude Max account
- MCP Calls
- [BUG] Unattended CCR routine sessions hang indefinitely on an unanswerable permission-mode prompt; run status disagrees across surfaces
- [MODEL] Scheduled /loop cron task accessed physical device (camera) without session consent after /clear
- [BUG] macOS: "Keychain is not writable" reported while security add-generic-password succeeds; login never persists (v2.1.276)
- Windows: Claude Desktop Terminal panel leaks one interactive ConPTY pwsh.exe (~85 MB) per Code session; ending the session does not terminate it, and closing the tab leaks its conhost.exe
- [BUG] Desktop local scheduled task: built-in AskUserQuestion is absent from the session tool surface, and the known workaround does not restore it
- EnterWorktree(path=...) to a valid but non-nested worktree reports success and moves Bash cwd, but the Write/Edit guard silently keeps the original binding
- [Bug] Anthropic API Error: Request Blocked by Safety Guidelines
- [FEATURE] Answer AskUserQuestion options directly from the push notification on iOS
- Grounding failure + prompt non-compliance: source ignored, invented workflow
- [Bug] Safeguard Blocking Legitimate Operations
- [Bug] Unexpected model downgrade to Opus 4.8 during C code generation
- [BUG] id: 152bd570-4955-41dc-a820-7a10daf822dd ----"CoworkVMService trava em loop de reinício com 'Access is denied'
- [BUG] New claude launches SIGKILLed (Code Signature Invalid) once the binary file has long-running sessions; a fresh copy of the same file runs
- [Bug] Claude ignores user instructions and performs unrelated actions
- Claude desktop app (com.anthropic.claudefordesktop) missing NSScreenCaptureUsageDescription/NSAppleEventsUsageDescription — blocks Screen Recording/Automation grants
- [Bug] Claude ignores explicit instruction to avoid rioplatense dialect in memory
- Docs: desktop app output style picker, per-session override and New style dialog are undocumented; per-session choice is invisible to CLI
- [BUG] VS Code extension ignores `bypassPermissions` mode; CLI honors it on same machine and config
- Slash command autocomplete can select the wrong command after fast typing
- Session .jsonl transcript silently deleted while session folder remains — deletion pattern doesn't match simple age-based retention
- [Bug] Update banner overlaps context banner on same line
- [BUG] Serena MCP tools (activate_project / list_memories) extremely slow or time out
- [Bug] Anthropic API Error: Missing server_tool_use block for tool_result
- Desktop app: Remote Control on a running session is dropped by every app relaunch (auto-update) and only restored on the next user message; deleting another session also switched it off
- [BUG] Cowork (web): near-identical assistant paragraphs re-rendered three times in one message block during/around auto-compaction
- I don't have a bug report to analyze. Could you please provide the bug report details that you'd like me to create a GitHub issue title for?
- [Bug] Claude Code downgrades to Opus instead of maintaining requested model
- [BUG] Claude Code: Memory/Standing Instructions Inconsistently Applied Over Multiple Sessions
- [BUG] Remote Control returns HTTP 403 on the in-session `/bridge` path, while `claude remote-control` works on the same machine and account
- [BUG] Claude Code deleted my source tree by running a recursive delete inside a OneDrive-synced folder, then told me to confirm the sync deletion dialog
- [Bug] Message flagged by safety system without clear reason during non-security tasks
- [Bug] GUI rendering issue in agent overall view
- [Bug] Control+Enter adds newline instead of sending message
- [Bug] Anthropic API Error: Message Flagged by Safety Filter (reasoning_extraction)
- Plugin cache: stale .in_use PID markers in orphaned version dirs cause intermittent resolution to outdated plugin content
- [BUG]
- [BUG] --resume re-sends the whole CLAUDE.md when only its path changed, not its content
- [FEATURE] Account-synced plugins have no update API: skills have upload-skill, plugins have nothing, so a hook fix cannot be shipped without a human clicking Accept
- /login reports "Login successful" but token is never saved: ENOTDIR rmdir on stale .storage-write.lock file (2.1.277)
- Claude Code deleted ~600GB: unprompted `rm -rf` on a substitution that resolved to a drive root
- [BUG] Claude in Chrome extension installed and signed in, but never connects to Claude desktop app
- [BUG] MCP connector calls hang indefinitely from scheduled-task sessions (work fine interactively)
- [Bug] Claude Code refuses to create WordPress login functionality
- [Bug] Session skill discovery doesn't reflect newly-added skills in already-enabled plugins
- Terminal CLI sessions never sync to claude.ai/code; desktop-app sessions on same account do
- [Bug] Overly broad safeguard false positive on legitimate IoT device API integration
- [Bug] Right sidebar TUI window cannot be closed with X button in iTerm2
- Browser pane never navigates: navigatePreview resolves context by `preview-{sessionId}` instead of serverId
- Supabase MCP fails with CONNECTION_CLOSED due to missing peer dependencies
- Agent asserts unverified causes as fact, acts on them, and persists them to memory
- [BUG] Remote Control: reattaching to a resumed conversation's bridge session never flushes history, so turns made before enabling are missing on claude.ai
- [BUG] Desktop Code tab: open editor does not reload files changed by Claude
- ~/.claude/sessions/<pid>.json's sessionId not updated on /clear (inconsistent with #88845)
- FileChanged hook stops firing after any cwd change in the session; CwdChanged watchPaths does not restore it (2.1.270)
- [BUG] Claude Desktop (Code tab, macOS): claude://code/new link shows its folder but starts the session in a scratch workspace
- Artifact version picker missing from Share menu
- [Bug] `/context all` no longer lists individual instruction files (regression, 2.1.277)
- [BUG] Local Network access silently denied for CLI due to Mach-O UUID collision with "Claude Code URL Handler.app" (macOS)
- [BUG] CoworkVM Background Services running issue preventing application startup
- [BUG] Claude Desktop + claude.ai: Projects grid leaves stale painted fragments after scrolling
- [BUG] PR chip stays in the bar after the PR is closed or merged, with no way to dismiss it
- Kitty graphics probe: iTerm2's i=less OK reply fails the response regex and is typed into the prompt as GOK
- Built-in terminal: stray character left at the start of the input line when the shell prompt contains a wide (emoji) glyph
- Cloud routines (CCR) silently fail, stall in PENDING for hours, or fire duplicates with no visible error
- [Bug] Anthropic API Error: Opus 5 (1M context) safeguards incorrectly flagging git configuration requests
- [BUG] Claude code in the web ui doesn't
- [Bug] URLs require multiple clicks to open in new renderer
- [Bug] Anthropic API Error: Claude 3.5 Sonnet safeguards blocking legitimate development work
- [BUG] 2.1.277: excludedCommands "every part must match" also drops single commands carrying a pre-subcommand flag (git -C, -c, --git-dir)
- [BUG] "Sync automatically" never creates a webhook on a personal (non-org) marketplace, despite correct GitHub App permissions
- [MODEL] [MODEL] Agent doesn't check current state before resuming work, redoing completed work
- [Feature Request] Make `/btw` forks configurable as parallel non-worker agents
- [BUG] [BUG] False positive safeguard trigger on safe coding conversation (reasoning_extraction)
- [BUG] False positive safeguard trigger on safe coding conversation (reasoning_extraction)
- [BUG] A blocking Stop hook's refused reply is still rendered to the user, so every refusal shows them the whole answer again — and two hooks can deadlock with no way out
- [Bug] Message display dimmed on initial submission before highlighting on macOS
- A `/goal` Stop-hook record written mid-turn splits an advisor server-tool call from its result, permanently bricking the session (400 loop)
- Desktop app: folder picker has no path input, and the diff pane is empty when the session root is not the edited git repo
- [BUG] A Stop hook's blocked reply is still shown to the user, so every retry reads as the assistant repeating itself
- iOS Simulator tool: touch/tap injection silently no-ops after Xcode 27 upgrade (Simulator.app replaced by headless DeviceHub)
- Worktree exit dialog offers "Remove worktree" for a worktree it will refuse to remove
- [Bug] Claude refusing legitimate security tasks due to overly strict safeguard filtering
- [BUG] Stacked skills: only the first is treated as user-invoked, so a disable-model-invocation skill in position 2+ is refused
- [Feature Request] Restore /low-priority command functionality
- [Bug] Inactive terminal window scrolling disabled
- Desktop app: folder picker "Recent" list capped at 8 (was 20+), and scheduled routines crowd out real projects
- [FEATURE] No way to switch model at the plan-approval step in the Desktop app's Code tab
- [BUG] Plugin account-upload is last-write-wins with no compare-and-swap, so concurrent sessions silently destroy each other's hook fixes
- [Feature Request] Implement confidence calibration and evidence tracking for root-cause analysis in coding workflows
- Regression: sessions >30 days old lost history again after forced re-sign-in (previously fixed in 1.34493.0)
- claude:// session deep link opens app but not the specific session (Remote Control "local" session)
- [Bug] Classifier over-triggering false positives in tool execution validation
- [MODEL] Treated a normal platform artifact as the bug symptom for ~2 hours without establishing a baseline
- [Bug] Concurrent GitHub edits cause agent confusion with file state
- [BUG] Exit hint prints `claude --resume "<title>"` for sessions the resume list hides (first user message is a slash command)
- [Bug] [cyber] safety classifier stopped 7+ turns across 3 sessions while building my own home camera system on my own LAN (Opus 5 and Fable 5.1; Opus 4.8 unaffected)
- [BUG] Single ~60 MB file upload writes one oversized transcript entry; session then never loads (blank window, Windows desktop)
- Desktop (Linux): Code tab session started without a project folder runs in a throwaway scratch workspace, silently disabling SessionStart hooks
- [Bug] All API requests reporting errors regardless of input
Resolved issues (250)
- I'm here to help you generate a GitHub issue title for Claude Code. Please share the bug report or feature request details, and I'll create an appropriate technical title for you.
- German translation: duplicate labels in session feedback prompt
- [Bug] Security flags raised for basic operations
- [Bug] Claude Code overriding user decisions from previous sessions
- [Bug] CLI ignores user input and makes decisions autonomously
- [Bug] Claude makes premature final decisions instead of presenting user choices
- I don't have access to the actual bug report content associated with that request ID. To generate an accurate GitHub issue title, I would need you to provide the details of the bug report, including: - What the issue is - Error messages (if any) - What be
- [Bug] False positive validation error on legitimate API requests
- I don't see a bug report in your message. Please provide the bug report details so I can generate an appropriate GitHub issue title.
- [Bug] DMCA §1201 False Positive in Security/Copyright Detection
- [Bug] Authorized requests incorrectly flagged as unauthorized
- I don't have a bug report to analyze. Please provide the actual bug report or error message you'd like me to create a GitHub issue title for, and I'll generate an appropriate title following the format specified.
- Claude Chrome extn fails
- [FEATURE] To avoid stale interactions: Store the timestamp of each interaction in Claude Code and sort them
- VS Code extension: group session history by project folder (Desktop parity) and allow cross-project session opening
- Feature: populate session_scope.displayName so Claude-in-Chrome tab groups are labeled per session
- [Feature Request] Auto-apply consistent treatment across related items without re-prompting
- [FEATURE] Proposal: Support a Shared `.agents/` Directory and Manifest-Based Discovery
- Feature request: inline WYSIWYG / live-preview editing for Markdown
- [Feature Request] Review pricing and feature parity with competitors
- [Feature Request] Add CSV export format support
- [Feature Request] Allow flag override for private psychometric assessment systems
- [FEATURE] - Token Refund and Pattern review
- [Feature Request] Refine content safety filtering to reduce false positives on legitimate cybersecurity certification study materials
- [Feature Request] Add support for building regression test suites with cross-tenant isolation validation
- [Feature Request] Add dismiss option for promotional banners
- Ctrl+scroll wheel zoom in the desktop app
- [FEATURE] Branch-level semantic eviction — collapse dead reasoning branches instead of whole-context compaction
- [FEATURE] allow pagedown and pageup for workflow agents detail page
- [FEATURE] I want to have back the possibility to filter by project in claude on the web.
- [FEATURE] Render images inline via sixel/kitty graphics protocol in terminal (CLI)
- I appreciate you reaching out, but I need to clarify a few things: 1. This interface is for reporting bugs or feature requests for **Claude Code** (Anthropic's CLI tool), not for general support requests or appeals. 2. I don't have context about any acco
- Feature request: Polish (and other language) localization of the CLI UI
- [FEATURE] Persist usage-banner dismissal across increments and multiple VSCode instances
- [Feature Request] Add persistent long-term memory integration with Cognee remember skill
- [FEATURE] Save a project as a reusable template for scaffolding future similar projects
- [FEATURE] Resume a self-filed GitHub issue's checklist as the session's task list
- [FEATURE] Detect low-quality manual commits and offer to clean up the message
- [FEATURE] Bulk-create/edit a scheduled Routine across multiple repositories at once
- [FEATURE] Surface stale open PRs that Claude Code opened on the user's behalf
- [FEATURE] Nudge when a session's working directory looks mismatched from what's being discussed
- [FEATURE] Consolidated digest for multiple scheduled Routines firing close together
- [FEATURE] Sidebar panel for surfacing open/critical questions in a conversation
- [Feature Request] Add first-class trust boundaries and untrusted zone isolation for prompt injection defense
- [Feature Request] Relax security policy for threat model generation on user's own software
- [Feature Request] Reduce false positive refusals for legitimate cybersecurity research queries in Claude 3.5 Sonnet
- [FEATURE] Per-agent frontmatter option to skip CLAUDE.md / git status injection for custom subagents
- Agentic session (subagent-driven development + systematic-debugging) fixed a bug but didn't catch that its own alerting had been silently broken for 5 days
- [FEATURE] Feature request: persist or let me replay the rotating spinner-tip hints
- [Feature Request] Add visual diff/comparison mode for iterative UI adjustments
- [Feature Request] Support for direct audio input from WhatsApp integration
- [Feature Request] Support for Level 4 Autonomy across portals, servers, and devices
- Mobile remote-control sessions: widget/visual tool output doesn't render
- [Feature Request] Computer use capability missing from Pro Team plan
- [Feature Request] Improve context retention and instruction adherence in multi-turn conversations
- [FEATURE] Add `app:compact` bindable keybinding action
- Feature request: enforce structured question events when an assistant turn requires user action (buried-ask invariant)
- Desktop app: keyboard shortcut to select working directory (keyboard-only workflow)
- [Feature Request] Allow credential input for local development servers
- [Feature Request] Auto-migrate deprecated Write() permission rules to Edit() in settings.json
- [FEATURE] Historical session file compression
- Allow opening the browser panel before/at the start of a new session
- Feature Request: `claude -c` should gracefully start a new session when no previous conversation exists
- Agent should check for native site features before building browser-automation workarounds
- [FEATURE]
- Request: Longer OAuth session duration for Max subscribers
- Feature request: OS-native masked-input prompt for pasting secrets (API keys/tokens) instead of chat text
- [FEATURE] Add interpolation for ANTHROPIC_CUSTOM_HEADERS
- [FEATURE] Mobile app Code tab: no Pinned sessions section, desktop/web session pins never surface on mobile
- [Feature Request] Add settings.json toggle to suppress footer hint row
- Reviewing image evidence is painful: delivered files lack open/copy-path affordances and can't be referenced back
- [Feature Request] Reduce over-flagging of benign academic content in safety filters
- [FEATURE] VS Code extension: option to disable Up/Down arrow keys switching between past sessions in chat input
- [FEATURE] Surface the teleport command ("Open in CLI") directly in the Slack session message
- Per-routine Runs panel shows only timestamp + generic "Completed" status, no per-run title
- API key scoping: differentiate application (API-only / Claude Code / Both)
- [Bug] Documentation: Clarify Claude Code policy on impersonation and content restrictions
- Routines sidebar shows only the fixed task name — no way to surface live per-run status
- [BUG] Desktop App disappears without warning and cannot reopen. Reinstall Required to work again
- [FEATURE] In Claude Code, store recent prompt history and allow for rerunning a recent prompt
- [Feature Request] Allow user override for medical content restrictions on capable models
- Status line: include permission_mode in the stdin payload (mode changes already trigger a refresh, but the field is missing)
- Feature request: UI localization (l10n) support for Claude Code VS Code extension
- [FEATURE] Warn before reviving a stale session - large cache-expired context re-billed on manual follow-ups AND automatic hook wake-ups
- [BUG] Workflow tool hard-rejects run_in_background with a bare 'unexpected parameter' error instead of stating the tool is background-only
- Workflow tool: persist each agent() call's label/phase in the agent's on-disk metadata at spawn time
- [FEATURE] Workspace boundary enforcement for local file writes — auto mode allows out-of-workspace Bash writes silently; recommended alternative (sandbox) unavailable on native Windows
- [FEATURE] Subcommand autocomplete for custom slash commands
- Feature request: share/attach conversation context across project folders (VS Code extension)
- Show remaining session/weekly usage on the Code tab dashboard alongside totals
- [FEATURE] Show Remote Control status in the agent view / session lists
- Expose an autoplay-policy control for the in-app Browser pane
- Resume-from-summary guard doesn't fire when the session prefix was invalidated by a CLI version change
- Adjudication step for auto-memory — batched promote/keep/discard review
- VS Code extension: tool OUT panels truncate narrative text instead of wrapping (re-filing #40901 / #41736 — the stale bot asked me to!)
- Claude Code repeatedly rediscovers solved problems (reopened – still relevant, closed only for inactivity)
- [FEATURE] Real-time plan usage/quota indicator in Cowork (5-hour & weekly limits)
- [FEATURE] Support multiple OTEL endpoints simultaneously for telemetry export
- [Feature Request] Let a Claude Design conversation be shared/transferred to a different Claude account than the one it was created under
- code-review: findings are adversarially verified, but the fix plan never is — a fix-plan design pass measurably breaks fix→regress cascades
- Feature request: session-only switch (`s` key) in /effort and /advisor pickers, matching /model
- [FEATURE] Documented interactive-vs-headless signal in the SessionStart hook input (so hooks can self-gate for claude -p / nested sessions)
- Agent overrides explicit, confirmed user instructions on low-stakes actions — and in one case the reason it gave was not the real one
- [Feature Request] Add support for F# Fable 5 zero-knowledge proof verification
- [FEATURE] Metrics - Count Azure DevOps PRs created with `az repos pr create`
- [Feature Request] Code Review performance optimization and setting recommendations
- [FEATURE] Allow customizing/swapping the Claude Code mascot icon
- Pricing feedback: Max 20x subscriber burning ~$3.9k/month in credits on premium-tier orchestration
- VS Code extension: Stop-hook-blocked draft and its rewrite both render as duplicate full responses
- [Feature Request] Clarify content policy categorization for agricultural operations and feedlot management queries
- [FEATURE] Managed Default for Sub Agent Model
- AskUserQuestion: render question bodies as markdown instead of all-bold text
- Feature request: local control API for external orchestrators (Codex Desktop parity) — blocks replacing Codex for some workflows
- [FEATURE] Expose pasted text content/boundaries to UserPromptSubmit hook (mirror imagePasteIds)
- [Feature Request] Add strict mode flag to error on unrecognized commands instead of starting session
- Feature request: keyboard shortcut to accept the assistant's suggested next action (e.g. Alt+Enter)
- [FEATURE] Expose Claude Design annotations/comments via MCP (read, reply, resolve)
- [Feature Request] Add context awareness for security audit workflows to reduce false safety flagging
- [Feature Request] Add option to disable safety guardrails for educational/scientific content
- [FEATURE] Reflect session state (awaiting permission / working / idle) in the browser tab favicon for Claude Code on the web
- [Feature Request] Add inline edit option to file modification prompts
- [Bug] Service unavailability or API error affecting paid users
- [BUG] Claude Code 2.1.243 immediately segfaults on Arch Linux x86_64
- CLAUDE_CODE_PROPAGATE_TRACEPARENT=1 does not inject traceparent header into outbound LLM requests
- [BUG] Stale "connector requires authentication" reminder persists after successful MCP tool calls (plugin-bundled connectors, desktop app)
- Hook commands resolve interpreters against the host app's live environment — a desktop-app relaunch can flip bash to /bin/bash 3.2 and hard-wedge every PreToolUse(Bash) hook
- [BUG] Claude Desktop crashes on Japanist (Oyayubi-Shift / 親指シフト) IME input after Electron 44 (Chromium 152) update
- Desktop/Web: sidebar group membership is device-local while Group by/Sort by syncs — make group assignment account state
- [FEATURE] Pool and dynamically reallocate unused enterprise AI credits at month end
- Android: Bluetooth keyboard input drops after exactly one character in chat input box
- Flagged as Dangerous via Fable
- [Bug] Multi-select prompts used incorrectly when single option is appropriate
- You've hit your session limit · resets 8:10pm (Europe/Rome)
- [BUG] Unexpected biology safeguards in Claude Opus 5
- [MODEL] OPUS 5 must go!!!
- [Bug] Security classifier malfunction
- [Bug] Claude Code chat does not respond to user messages
- [Bug] Claude Code building changes in temporary directory and using rsync instead of direct file operations
- [Bug] MCP server validation bypassed when remote-settings is empty
- cant scan with mcp in session idk why
- I need the bug report content to generate an appropriate GitHub issue title. Could you please provide the bug report details you'd like me to create a title for?
- I don't have access to the bug report details for request ID `req_011CewrLoZnRaugrk1YeBz9T`. To generate an appropriate GitHub issue title, I would need you to provide the actual bug report content, including: - What the issue is - Error messages (if any)
- I need more information to generate an appropriate GitHub issue title. Could you please provide: 1. What error message or behavior did you encounter? 2. What command or action were you trying to perform? 3. What was the exact error or restriction that occ
- [FEATURE] Please make the visibility of `/cost` / `/usage` configurable so that it can be retained in the .jsonl logs
- Sumiu Funcionalidade da extensão Claude no VSCode
- [FEATURE] Make background session idle timeout configurable
- Post-mortem weekend update: weekend three of fuck-all got done with Fable
- [FEATURE] Desktop app: 'folder missing' section + bulk re-link for sessions whose project folder no longer exists
- [FEATURE] Sessions should follow a renamed/moved project folder (persistent project identity)
- Feature request: harness-anchored trust for injected messages (operator identity, orchestrator amendments, signed commissions)
- Max plans: include a recurring ultrareview allotment instead of 3 lifetime free runs
- [FEATURE] Desktop file viewer: allow editing `.md` files in place, as code files already allow
- [FEATURE] Make the Ctrl+S prompt stash a LIFO stack (git-stash-style push/pop) instead of a single slot
- [Feature Request] Allow user override of automatic model downgrade for security tasks
- Claude Code on the web: no way to supply plugin `userConfig` (secrets) to cloud sessions
- Desktop/Cowork document review: batch multiple comments and edits on a rendered file into one submission
- [FEATURE] PostRewind hook event — hooks that manage external state can't observe /rewind (re-files stale-closed #18858)
- Feature request: configurable output pacing / scroll throttle for streamed responses
- [FEATURE] Show each session's working directory in the desktop sidebar
- [FEATURE] Calibrated trust/verification score on diffs and PRs before merge
- [FEATURE] Link production incidents (Sentry/Datadog) back to the originating Claude Code session/commit
- Session groups sort oldest-first with no option to reverse. In a group with 200+ sessions the one I want is always at the bottom. Please add a persistent sort toggle: newest first (by last active) or oldest first. Pinning and manual reorder don't scale.
- Feature Request: Add tool response diffing/delta for MCP tools to reduce context window usage
- [MODEL] Opus 5 (1M) reports itself "out of context" at 28-50% usage and stops working; Opus 4.8 does not
- [FEATURE] VS Code extension: make user-message blocks visually distinguishable (themable color / configurable style)
- [Bug] Advisor state not exposed in status line JSON; workaround produces false indicators
- Permission dialog / confirm UI text is not localized (stays in English even with language: "japanese" set)
- [FEATURE] GitHub Integration connector: expose GitHub Projects v2 (boards) tools
- Cannot resume a normal session into an agent already pinned to AgentView; should redirect to AgentView instead
- [FEATURE] Desktop: clone a Git repo from a URL when starting a new session (no terminal)
- [Feature Request] Clarify Claude model version and performance consistency in Claude Code
- Desktop: keyboard-navigable session list with synced full-description detail window (hybrid of CLI agent view + Desktop sidebar)
- [FEATURE] QA engineer running Appium suites, wants to watch the device without leaving the app
- [FEATURE] Selectable ghost-text prompt suggestions — cycle through N, not just one
- Feature request: built-in git-worktree auto-sync for --spawn worktree multi-session workflows (working hook pattern included)
- Cross-session messaging between Cowork and Code sessions (currently code-to-code only)
- [Feature Request] Custom agent tags and user-defined view sections for agent management
- [Feature Request] Allow configuration of insecure default credentials for development environments
- [Feature Request] Add account-level safeguard configuration or allowlist for legitimate use cases
- [Feature Request] Pre-configure WSL environments for Desktop sessions (wslConfigs / deep link / trust via settings)
- Feature: show `claude rc` remote control sessions in the agents terminal view
- Feature: run `claude rc` as a background daemon (no open terminal, auto-restart on reboot, start/status/stop control)
- Feature: let `claude rc` expose all subfolders of a parent directory as selectable remote sessions
- Make model used for /compact command configurable
- Enshrine — a way to keep the conversations that mattered (distinct from archive)
- [FEATURE] Cowork: surface local vs. cloud execution mode in the composer, and say when local context isn't loaded
- [Feature Request] Rethink shell expansion permission prompts - resolve or skip, don't prompt on undecidable strings
- [FEATURE] Custom agent definitions can rot silently — no native way for a required verifier to detect its own anchor is stale before it stamps work as verified
- [Feature Request] Support Fable 5 language in plan mode for security audits
- Remote Control: register a session without the 'session ready' push (fleet/automation use)
- [FEATURE] Claude in Chrome: connected browser can be driven cross-machine with no reliable device identification
- [FEATURE] Component Analysis feedback for CLAUDE.md
- permissions.deny on an absolute credential path silently matches nothing (single leading / is project-relative) — no warning
- [BUG] PreToolUse hook `if: "Bash(...)"` still false-positives on `$()` command substitution in 2.1.202
- [Feature Request] Implement direct coordination workflow between Claude Design and Claude Code for asset handoff validation
- [FEATURE] Suggest targeted logging to collapse the hypothesis space
- VS Code extension: auto-open/pin the file Claude is reading or editing in a side editor group
- Support multiple cron expressions (or multiple Schedule triggers) for a single task
- [FEATURE] Permitir abrir enlaces file:/// con el explorador del sistema operativo (no solo el panel interno)
- [FEATURE] plugin install --config: support reading values from stdin to keep secrets out of process argv
- Desktop app: highlight color for open sessions in the sidebar is too subtle and not customizable
- [Feature Request] Support image upload with multiple size specifications
- [FEATURE] Credential guardrail should recognise operator-owned secret moves and stay consistent in-session
- Claude Code recommended Hetzner EU server at ~€4/month — pricing is outdated, cheapest available is now €12.42/month
- [FEATURE] Expose subscription billing-cycle / renewal date in statusline JSON (or via CLI)
- [FEATURE] Support disable-model-invocation in agent frontmatter (plugin agents load descriptions unconditionally)
- [Feature Request] Display task numbers in descriptions for chat reference
- [Feature Request] Support security-focused tasks in system prompts without safeguard restrictions
- [Feature Request] Harden published Artifacts against "trusted-domain" social engineering / malware distribution
- [Feature Request] Allow Claude Code usage for authorized security research and bug bounty programs
- Named headless (-p) sessions should be visible in the /resume picker and resumable by name interactively
- Feature request: show the model powering each subagent task in the task display and /tasks list
- [Feature Request] Add option to display per-message timestamps in interactive TUI
- [ENHANCEMENT] Display MCP resource title/name in @-mention chip while retaining URI value
- [BUG] Move an existing task in cowork to a project
- VS Code extension shows full subagent transcript inline; no setting to collapse it (CLI/desktop show a summary)
- [Feature Request] Add option to disable Fable model safeguards for hardened systems
- [Feature Request] Support OWASP Top-10 security scanning for custom web applications
- [Feature Request] Let a Claude Design session talk directly to its paired Claude Code session — right now the handoff is manual
- [Feature Request] Reduce restrictive feature gates and access limitations
- Mid-turn user messages lack reply-target identity — crossed replies are unattributable in orchestration-heavy sessions
- Feature: expose background Bash/Monitor task liveness to hooks (Stop/SessionEnd) via a stable, documented signal
- [MODEL] No exemption for local/private-network origins on the browser JS-execution permission gate
- MCP tool approval dialog: array/object params render as one compact JSON line, not pretty-printed
- [Feature Request] Allow model switching between Claude Opus and Fable for security testing
- Modifier-drag (hold Control) to force file-path insertion for all dropped files, including images
- [FEATURE] Allow custom filenames/labels for pasted (Ctrl+V) screenshots
- [FEATURE] Claude Code Preview panel for remote sessions
- Show repository/project name in the agent (background jobs) view
- [FEATURE] Support `updatedToolOutput` in `PostToolUseFailure`
- Claude replies even when explicitly told 'do not reply now'
- transparency what the limits actually are!
- [FEATURE] Render hook system messages with user-facing labels
- Claude Code (VSCode extension) cant read virtual/on-memory Sailpoint Extension
- Expose the current turn's assistant text to PreToolUse hooks (validate "text before tool_use")
- [FEATURE] let the harness surface Claude's just-authored text for approval without a second model pass, especially on remote/mobile(token cost)
- [Bug] Browser tabs lack session-specific titles for tab differentiation
- [FEATURE] Sidebar: option to hide sessions already open in an editor group
- [Feature Request] Implement context-aware model switching to avoid re-reading conversation history and optimize token usage
- Built-in browser can't load new Microsoft Teams web app
- Support per-model effort level defaults
- Feature: Auto-save large pasted text to a scratchpad file (like Copilot CLI)
- [Feature Request] Allow model selection for max plan users instead of forced Opus 4.8 upgrade
- [Feature Request] Auto-cleanup empty .claude/.cc-writes/ directories after execution
- VS Code extension: prompt input focus ring ignores theme `focusBorder` (hardcoded orange)
- [Feature Request] Allow code review of user's own source code without flagging
- No mechanism to require checkpoint/approval before large-scope generative rewrites of user content
- Provide a structured DAG view for orchestrator/subagent workflows
- Desktop app: no discoverable way to switch project/working folder for a new session
- [FEATURE]