Digest / September 8, 2026
Claude Code Issues Digest — September 8, 2026
200 new issues filed ·
510 resolved in anthropics/claude-code (UTC day).
New issues (200)
- I need more information to generate an appropriate GitHub issue title. Please provide the bug report details, such as: - What error message did you encounter? - What were you trying to do? - What was the unexpected behavior? - What system/environment are
- [Bug] Case-sensitive session name matching in /resume command
- [BUG] Windows icon renders as wrong/blurry variant at DPI scaling >100%
- [BUG] Ctrl+G external editor receives no keyboard input inside ConPTY-based terminal multiplexers on Windows (refiling #73301)
- Local MCP servers fail with 'Not ready after 60 seconds' even though the remote connection succeeds (shared-pool readiness bug)
- Desktop app: file pane lost its copy button; ⌘A selects the chat too
- [BUG] Remote Control fails with "Remote Control server rejected the request (HTTP 403)" on Pro plan (Windows, Claude Desktop)
- [BUG] Linked worktree loads a project-scope plugin from the first same-repository row in installed_plugins.json, not the worktree's own row
- Model asserts and recommends without consulting the local record it wrote itself; failures are consistently self-favourable (related to #92505, #92512)
- [Bug] Claude Code not following user directions
- Background daemon forces full re-auth after PC shutdown instead of retrying token refresh
- [Bug] Anthropic API Error: Claude 3.5 Sonnet Cyber Safeguards Blocking Legitimate Requests
- [Bug] False Positive Detection in Code Analysis
- [Bug] Claude fails to consider object rotations in 3D packing optimization
- I don't have a bug report to analyze. Please provide the bug report details so I can generate an appropriate GitHub issue title.
- Plugin skills with disable-model-invocation: true are hidden from the invocable skills list
- [BUG] Windows 11: "Code" toggle unclickable when sidebar is in floating mode, works when docked
- Windows: Shift+Enter indistinguishable from Enter (modifiers dropped by libuv console-to-VT translation)
- Pin the last user prompt in the top bar (like Grok Build)
- Routines list empty (API + Web UI) and permission push notifications not arriving, execution still works
- Option to swap Enter and Cmd+Enter in the composer (Enter = newline, Cmd+Enter = send)
- [Bug] Claude Code ignores explicit instructions and makes assumptions without database verification
- [BUG] OTel token.usage / cost.usage silently stop recording main-loop turns mid-session (3% of actual captured)
- [Bug] Feedback submission incorrectly reduces prompt cache usage in active session
- [Bug] Unauthorized branch and worktree creation during operations
- Reproducible Windows BSOD — STOP 0x50 in bindflt.sys, always in claude.exe context
- [Feature Request] Fast mode spend is invisible until it runs out: no line item, no in-session balance, no warning before exhaustion
- Trackpad scrolling triggers Up/Down command history after /clear
- [Bug] Model falsely attributes user decisions to operator in autonomous agent workflows
- I appreciate you reaching out, but I'm unable to generate a GitHub issue title from this message because: 1. The text is in Korean and appears to be an expression of frustration rather than a structured bug report 2. There's no clear technical problem des
- macOS: "access data from other apps" (SystemPolicyAppData) prompt recurs every session on the SAME version — grant stored session-scoped (desktop-bundled worker)
- [BUG] Claude Desktop for Mac freezes every 7-10 minutes (EventEmitter listener leak)
- [Bug] Diff view assumes current directory is git root, fails in parent folder of multiple git repositories
- claude.ai connectors mount under connection UUIDs in Desktop local sessions, so `mcp__claude_ai_<name>__*` permission rules never match there (re-raising #77598, closed as stale)
- AskUserQuestion silently discards typed free-text answer on single-select when a different option is selected
- Session hangs indefinitely — queued prompt never dispatched (reproduced 5x, 2 unrelated projects)
- [Bug] Intermittent API failure causing tool execution failures
- Bash 工具在多次呼叫間悄悄重置工作目錄，導致破壞性 git 指令跑錯 repo
- [Bug] Connection timeout in restricted network environment
- [Bug] Desktop app: after adding a GitHub plugin marketplace, it can't be found or installed
- Classic renderer: mouse wheel and click-drag selection dead on Ptyxis/VTE 0.80 (Wayland); CLAUDE_CODE_DISABLE_MOUSE=1 has no effect
- [Feature Request] Changes to agent are prevented when running agent in a separate session
- [Bug] Claude Code generates incorrect code suggestions sequentially
- Scheduled routines: tool calls wait for a permission prompt nobody can answer, and the run still reports success (follow-up to #83166)
- [BUG] parse-frontmatter.sh returns body content as frontmatter when the markdown body contains ---
- [Bug] Preview pane freezes during inline text editing in Artifact canvas on Chrome/Mac
- [Bug] Credit auto-consumption fails to pause when subscription limit is reached during subagent execution
- [BUG] Workspace trust not persisted across sessions (Code tab, macOS 15.5)
- [Bug] Mobile Remote Control lacks subagent monitoring, manual cancellation, and query visibility
- Feature request: spoken responses (TTS with Claude's voice) in Claude Code
- [Feature Request] Add persistent state management and monitoring for multi-agent orchestration workflows
- [BUG] [BUG] "Code" tab missing entirely from Claude Desktop app (macOS, Pro plan)
- I don't see a bug report in your message. Could you please provide the bug report details so I can generate an appropriate GitHub issue title?
- [Bug] Overly broad safety filter flags legitimate telecom debugging as "Cyber"
- Browser pane webview unloads (blank URL bar) after idle, independent of backing server
- [BUG] VS Code extension: chat file links are never wired to the existing open_file bridge
- [BUG] active sessions lost when doing window reload on Antigravity IDE
- [BUG] sessions started in mobile app on a remote server cannot be picked up in cli on remote server
- Enable Credentials vault feature for my account
- [BUG] Switching between accounts in Claude Desktop appears to pause running tasks on the other account
- [BUG] Claude Code Agent Force-Kill Proses Excel Tanpa Verifikasi — Kehilangan Data & Pemborosan Token
- Channels: allow a persistent/settings-level channel opt-in so already-running sessions can receive pushed events
- [FEATURE] `.agentsignore`: gitignore-syntax file for paths agents must never access
- [Bug] Bypass all permissions setting not respected during operations
- [BUG] [VS Code] "New session" ignores the requesting panel's editor group and opens a new locked group unless some group holds only Claude tabs
- [Bug] Game design task terminated by safety classifier
- Custom subagents that declare Bash in frontmatter never receive it (no config can restore it)
- [Feature Request] Add custom named session collections/groups
- [FEATURE] Configurable audio transcription endpoint for dictation in Cowork on 3P
- [MODEL]
- [Feature Request] Add clipboard image paste support for direct screenshot insertion
- [BUG] Claude Code Desktop: session transcripts silently become permanently unavailable (cliSessionId nulled, no local recovery path) — follow-up to anthropics/claude-code#79044
- [Bug] Rate limit are too low
- Queued message that starts with a dragged file path ("/Users/…") is never delivered mid-turn; it waits for the whole turn to end
- [FEATURE]
- Desktop app: multi-agent Workflow run white-screens the app and starves an 8 GB machine; in-flight subagent work is lost
- VSCode extension has no sound/notification path — preferredNotifChannel is a no-op there
- Desktop app: file cards and file-preview panel can't reliably show an existing file
- Text typed in the permission-deny dialog is lost when denying one of several parallel Agent tool calls (VS Code extension, 2.1.263)
- [Cowork] Scheduled task runs deny mcp__workspace__bash and mcp__workspace__web_fetch by permission rule since the Sept 2 desktop update (interactive tasks unaffected)
- You've hit your weekly limit · resets Sep 15 at 7pm
- [BUG] HTTP MCP client appears to send an unexpected MCP-Protocol-Version header after initialize, causing spec-compliant servers to reject all follow-up requests
- [BUG] VS Code extension spawns two `claude` CLI processes per launch, which collide on `~/.claude.json`
- Option to hide or disable auxiliary/helper sessions (title-generation, classification) from session history
- WSL2 repeatedly OOM-crashes (systemd killed) during multi-project Claude Code workflow/skill-generation task
- [BUG] MACOS Keychain: argv fallback for oversized OAuth credential payload (>4KB) itself fails silently ("pbt" error) during MCP OAuth login
- Auto-mode classifier: hard blocks are indistinguishable from soft blocks, and every remediation requires a keyboard the user may not have
- [Bug] Duplicate '/design' route options in CLI menu
- [Bug] Physics simulation error during game development
- [BUG] SSH connection fails with "no route to host" (EHOSTUNREACH) despite identical ssh command working outside the app
- [Bug] /code-review --comment creates separate reviews instead of single review with summary
- Asserts causal explanations without verifying; treats silently-failed commands as evidence of absence
- [Bug] False positive cyber security flag on local inference server connections
- [BUG] Native Windows: startup waits ~4.5 s on three embedded-ripgrep calls that each re-launch claude.exe; agent-view sessions take ~10 s to open (USE_BUILTIN_RIPGREP=0 halves it)
- [FEATURE] Google Sheets Edit
- [BUG] The second `/compact` in a process parents the transcript's continuation before the first compaction, orphaning every message in between
- [BUG] Claude Desktop: knowledge-work-plugins installed on the account have no remove or disable control on any surface
- I appreciate you reaching out, but I need to clarify that this appears to be a support request rather than a bug report for Claude Code. However, if you're experiencing technical issues with Claude Code specifically while using it for OSCP lab work, pleas
- [Feature Request] Add configurable session timeout settings for organization accounts
- [FEATURE] Expose a session's context window and cost to local tools when the session has no status line (VS Code extension, SDK)
- [Feature Request] Add passkey and password authentication options to reduce login friction
- [BUG] TUI renumbers the user's own ordered list when echoing it — every item after the first loses its typed number (regression in 2.1.234)
- [BUG] Desktop sidebar: sessions from GitLab subgroup repos never group by project on a second device (always land in "Other")
- [FEATURE] Iterative or Q&A Mode
- MCP stdio server tool-list stale across new sessions and full reboot (Claude Desktop, Code tab)
- Usage stats: the "Lord of the Rings" comparison is a vanity metric - no baseline, no breakdown, no action
- [BUG] Files viewed in Claude Desktop do not consistently update when changed.
- Desktop app adopts the server placeholder "Untitled session" as the final auto-title and never retries
- [BUG] Fable 5.1 regression in honesty and alignment making Claude Code unreliable partner
- [BUG] German Read Aloud ignores voice selection on macOS and web; works on iPhone
- [BUG] Claude Code in VS Code Agent hangs indefinitely and requires “continue” to resume
- [BUG] Cmd+C is interpreted as Cmd+W (close view) with the Ergol layout
- Statusline OSC 8 hyperlinks render as plain text in sessions created from agents view
- EnterWorktree(path=...) reports success but wedges/refuses all subsequent Bash calls for a worktree-isolated subagent
- Monitor tool: a curl+parse polling loop can report a false terminal "DONE" state on the first iteration while the real status is still pending
- A worktree-isolated agent's own sub-dispatched Agent calls inherit its worktree/repo, regardless of the declared target repo
- [DOCS]
- [FEATURE]
- [BUG] Desktop: Run button not offered on powershell/pwsh/ps1 code blocks, though Run itself executes PowerShell on Windows
- Untrusted-trigger PoC 1788882099
- Test issue
- [BUG] Dispatch: reasoning_extraction safeguard fires on bare 'hello' since Sept 1
- [Bug] Cyber Verification Program: Still flagged for basic tasks despite approval
- [FEATURE] /context: option to display estimated dollar cost instead of/alongside tokens
- You've hit your weekly limit · resets Sep 12 at 12pm (America/Los_Angeles)
- [Bug] Anthropic API Error: Request Rejected for Safety Policy Violation
- [BUG] "Carry" and "Bare" over-usage. Annoying, incorrect, and destroying all content generated by Claude Code.
- [FEATURE] /context: minimum-token display threshold to hide negligible entries (0-token MCP tools, <20-token skills)
- [Bug] Local sessions fail to load in Claude Code tab with "Not found" assets-proxy error
- Account reactivated without instruction after cancellation, then charged twice; further charge attempts continuing
- [FEATURE] Let --resume continue a Claude app conversation in Claude Code
- [BUG] Cowork: expose execution mode (Local/Remote) as a visible per-project setting, and document what leaves the machine
- [Bug] Anthropic API Error: Overly broad safeguards flagging legitimate game design content with BIO classifier
- [BUG] claudeCode.usePythonEnvironment silently no-ops when claudeCode.useTerminal is enabled
- [BUG] Python extension's venv activation is typed into Claude's prompt input instead of a shell
- [Bug] Code review stalls or fails on medium complexity tasks
- [Bug] False Positive Safety Flag on Non-Code Content
- Hooks: no way to intercept or suppress teammate idle_notification messages delivered to the lead (UserPromptSubmit does not fire on them)
- Artifact live mermaid renderer (pinned to 11.16.1) lays out complex flowcharts incorrectly — edges bundle at subgraph borders, implying wrong dependencies
- Injected commit/PR attribution instruction overrides the user's CLAUDE.md trailer rule — attribution should be defined by the project, not the CLI
- [Bug] Security regression false positive during game development
- [Bug] Scheduled cloud task failed to execute at configured time and did not send notification
- [BUG] Each sandboxed Bash tool call blocks the event loop ~8.4s (fixed cost, independent of command); terminal input freezes for the duration
- [Bug] Scheduled tasks fail to complete with Slack message output
- [FEATURE]
- [BUG] No non-destructive way to steer an in-flight task: correction sits unread for minutes, and interrupt terminates running bash processes and subagents
- MCP (stdio): one boolean property schema silently drops ALL tools of the server (SDK 0.1.77; stdio twin of #88049, re-filing of #82949)
- [BUG] sandbox.credentials.files protection can be bypassed on first access to a not-yet-existing deny-listed path
- Skill export ZIP is incompatible with Teams skill upload
- [BUG] Continuing from a session limit hit causes 45% of immediate quota usage (Fable 5.1)
- [BUG] Microsoft 365 connector unusable in Claude Code (works in Claude Desktop)
- [Feature Request] Make Opus 4.1 fallback behavior less aggressive
- [BUG] Weekly usage jumped from 0% to ~60% after two short conversations; "Approaching weekly limit" banner contradicts displayed 27%/13%
- [FEATURE] Dynamic real-time voice switching in voice mode (spoken command + programmable per-session voices)
- [BUG] Windows: `.claude.json` is saved as a whole-file read-modify-write behind a best-effort lock that is abandoned after 15 s — entries written by one process disappear again, trust flags included
- [Feature Request] Add bidirectional voice conversation mode for real-time interaction
- [BUG] [BUG] Routines (scheduled tasks) stuck in PENDING forever — sessions never execute
- [BUG] IS_DEMO still suppresses the workspace trust prompt in 2.1.263 — all hooks, statusLine and project permissions silently disabled, and /doctor has no trust check
- [MODEL] "Voseo rioplatense" como idioma por defecto en vez de Español Neutro
- [BUG] #44042 repost
- [Bug] Claude ignores explicit tool version upgrade instructions and reverts to stale versions
- [Bug] Unauthorized plan upgrade without user consent
- [BUG] Claude Code 2.1.263 crashes with Bun segmentation fault on Windows 11 — reproducible on trivial prompts
- [Bug] False positive content moderation flag for legitimate game modding
- `claude plugin enable|disable --scope project` rewrites .claude/settings.json and empties `footerLinksRegexes`
- [Bug] Line duplication and content loss on scroll up
- [BUG] Scheduled task dispatch acknowledged by renderer but no session starts; 'Cleared stale pending dispatch' after 12 min, lastRunAt stamped anyway (3 of 18 tasks lost overnight)
- [BUG] Cowork hangs after first message on Windows — clean VM boot confirmed, matches #26805/#26921/#44776/#59532 cluster
- Separate notification controls for scheduled task runs and normal chat sessions
- All routes to claude.ai redirect to onboarding [BUG]
- [BUG] Bedrock: /model picker hides all models in a family when default has a [1m] suffix
- [Bug] Legitimate debugging forensics incorrectly flagged as illegitimate
- Custom theme diffAdded/diffRemoved overrides still ignored in file-diff view (v2.1.236); light-ansi base also renders single-word diffs invisible
- Claude Design: false 'hit your limit' pause at 95% weekly on Max 20x; Resume succeeds after ~30 clicks
- Desktop + WSL: "+" on a project row opens the generic New view instead of starting a session in that project (Local works)
- [BUG] "tell Claude what to do next" instructions don't reach Claude if a bash command is approved but fails
- Passing --add-dir suppresses the initial positional prompt; the session opens and never starts a turn
- [BUG] Typo in workspace trust warning: "untusted" should be "untrusted"
- New session hangs indefinitely when cwd is HOME with no git repo - unscoped rg scan hits TCC-denied paths
- [Bug] Claude Desktop main window vanishes silently and stealth update fails to relaunch app
- Windows: synchronous command hooks wait for persistent descendants
- [Bug] Anthropic API Error: Claude Opus 4.8 Cybersecurity Safeguards Blocking Legitimate Requests
- [FEATURE] Supported per-generation admission: max-turns=1 still permits output-limit recovery requests
- [FEATURE] Show session color in Session Picker
- [BUG] Prompt cache for resumed session history intermittently fails to hit on `resume`, independent of idle time or TTL
- I need a bug report or feature request description to generate a GitHub issue title. Please provide details about the issue you're experiencing with Claude Code, such as: - What error or problem occurred? - What were you trying to do? - What was the expec
- I need more information about the actual bug or issue you're experiencing with Claude Code to generate an appropriate GitHub issue title. Could you please provide: 1. What error or problem are you encountering? 2. What were you trying to do when it happen
- [BUG] Blank screen / rendering freeze when a slash command auto-expands into a long input
- [BUG] Windows: app won't launch — "Another program is currently using this file" — cowork-svc.exe holds its own image inside the MSIX package
- Keyboard focus for the diff/review panel (no way to "enter" it)
- Desktop app: "Add a folder" prompt triggers logout, clears conversation/project list from UI (data confirmed intact on account via claude.ai web)
- showThinkingSummaries: false is ignored — thinking blocks still render (desktop app, Code tab)
- [Bug] Anthropic API Error: Content Blocked by Safety Filter on Legitimate Low-Level Debugging Code
- Claude infers permission to touch an explicitly forbidden folder from an unrelated general statement
- Cowork (claude.ai web): a single uploaded image is delivered to the model twice, causing it to assert the user sent two
- [BUG] Auto-compaction re-injects the CLAUDE.md/MEMORY.md copy from the last user prompt, not the on-disk file; disk re-read only happens at the next prompt
- [FEATURE] Persistent subagent inspector in the TUI — reopen any subagent's transcript, message it directly, edit its metadata
- [Bug] Agent loses context and makes repeated mistakes during extended sessions
- [BUG] Agent task notification reports status=completed for an agent still waiting on its own background task
- [Feature Request] Implement agent spawn validation to reduce unnecessary token usage
Resolved issues (510)
- I appreciate you reaching out, but I'm unable to generate a GitHub issue title from this message because: 1. The text is in Korean and appears to be an expression of frustration rather than a structured bug report 2. There's no clear technical problem des
- [BUG] Cowork remote-surface OTel events missing user identity attributes (user.email, organization.id)
- claude --version segfaults (SIGSEGV) immediately on 2.1.243 — 2.1.241 works fine on same system
- [BUG] [BUG] Accumulated notification/task messages in sidebar don't dismiss when tapping X
- [BUG] Claude Desktop crashes mid-session during active use, no quit event or system crash report, only a Crashpad dump
- [BUG]
- Claude refused direct instruction to include user-provided content in a draft
- [Bug] Anthropic API Error: Fable Safeguard Triggered Unexpectedly
- [Bug] Subagents in ultracode mode ignore window size limits
- [Bug] Claude ignores user-defined rules after reading documentation
- [BUG] The app close when i ask for something and then never opens again
- [Bug] False positive safeguard violation on Fable 5 code
- [Bug] Claude Code ignores provided tools and skills without notification
- [Bug] Misfire safety guard message not displaying correctly
- [BUG] Suspeita de corrupção de disco
- [Bug] API Subsystem Flagged - Unclear Error
- Hey I am building a swarm brain AI imagination method and...
- I'm ready to generate GitHub issue titles for Claude Code bug reports. Please provide the bug report details and I'll create a concise, technical issue title following the guidelines you've outlined.
- [Bug] Cannot switch models to Opus
- [Bug] Claude Code flagged for OpenWrt router firmware flashing operations
- False positive: Fable 5 safeguards flag routine Spanish text-editing prompts, forcing switch to Opus 5
- [Bug] Claude Code auto-downgrrades to Opus for local code audit without user consent
- [Fable 5] Agent treated a scoping question as authorization for a multi-hour autonomous run, ignoring an explicit repo instruction to confirm first
- [Bug] Extension fails to initialize in Chrome incognito mode
- Session link is written into git history by default with no reliable opt-out
- Subagents silently served a different model than requested when Fable overage consent has not been granted on that machine
- Claude in Chrome extension never connects (tabs_context_mcp always returns 'not connected')
- [Bug] Messages incorrectly flagged as cybersecurity policy violation
- Scheduled-task context injected at session start is sometimes not acted upon (idle greeting instead of executing)
- Fable 5 guardrails false-positive on routine sysadmin work → silent downgrade to Opus 5, with no indication of what was flagged
- Claude refuses file deletions citing a "hard boundary" that doesn't exist — argues past direct owner authorization; inconsistent across sessions on the same machine
- [BUG] Idle VSCode sessions leave native-binary/claude child spinning at 100% CPU for days
- claude mcp login (and bare claude) fail with "stdin isn't a terminal" in a genuinely interactive Windows Terminal + PowerShell 7 session
- [Bug] Claude fails to accurately extract items from lists and generates hallucinated content
- [BUG]
- [BUG] claude.ai Gmail connector: read/compose calls fail with 'insufficient authentication scopes', and in-session /mcp reconnect does not re-grant scopes or update session state (Windows, v2.1.200)
- [Bug] Safeguards trigger incorrectly on cybersecurity marketing context
- [Bug] False positive security flag during defensive code audit of own repository
- [MODEL] Fable 5 safeguards false-positive: routine SSH administration of own infrastructure consistently reroutes sessions to Opus
- [Bug] Anthropic API Error: Fable 5 Safety Filter False Positives on Legitimate DevOps Operations
- [Bug] Agent evaluates PR against stale base, confounds variables in before/after comparison, produces false conflicts
- [Bug] File edits hang indefinitely during processing
- [Bug] Abnormally crazy high credits usage sometimes. Not Fable
- [Bug] Agent incorrectly interprets injected "user " prefixed messages as user input
- [Bug] Claude loses context and forgets hardcoded rules during multi-turn conversations
- [Bug] False positive security flag triggers unnecessary model downgrade mid-session
- [Bug] Model hallucinating requirements and refusing valid production deployment requests
- [Bug] Claude 3.5 Opus model producing verbose responses without direct answers
- [FEATURE] Native multi-provider model routing in settings.json (per-subagent / per-model-alias provider selection)
- Boolean JSON Schema at a named property breaks MCP proxy construction — all tools of the server are dropped (#50194 still reproduces, wrong root cause)
- [Bug] Model crashes during security testing without clear error
- Fable 5's safeguards flagged this message...
- [Bug] Anthropic API Error: Cyber safeguards false positive blocking legitimate automation workflows
- [Feature Request] Auto-continue on transient API errors without manual intervention
- [BUG] M`/plugin marketplace update` + `/reload-plugins` does not pick up pushed plugin changes
- Always-loaded CLAUDE.md rules & persistent memory don't reliably prevent judgment-type reasoning errors
- [Bug] Tool refuses to read and display clone instructions, generates hallucinated responses instead
- Wasted Anthropic Credits
- Opus 5: countable thresholds in CLAUDE.md are read as removing discretion (with reproduction + model's own explanation)
- [Bug] Claude Code not retaining context of previously identified security issues
- [Bug] Model performance degradation in chat session
- Allow opening a new empty chat directly from the Agents view
- [Bug] API key exposed in filtered output logs
- [Bug] Model switching when discussing neuroscience context in reasoning tasks
- [Bug] Anthropic API Error: False positive safety filter on legitimate safety/moderation product context
- [Bug] Safeguard swap misattributes rejection to user and may terminate background tasks during security work
- [BUG] Constant requesting for access to everything. Claude code has become useless as a coding tool.
- Session/agent management should be opt-in: plain CLI sessions get entangled with background-job topology
- Permission classifier blocks previously-allowed commands non-deterministically, citing 'earlier conversation content'
- [Bug] Bash commands with ! operator execute in wrong directory in git worktree
- [BUG] Agent tool model: "opus" override spawns a subagent that self-reports as Sonnet 5
- Claude-in-Chrome extension appears stuck on 1.0.84, contradicts "fixed in 1.1.9310" note on #39707
- Workflow: schema-forced custom agentType can end its turn with prose and stall silently (no per-generation watchdog)
- [BUG] Max 5x gift subscription redeemed but not provisioned — Fin AI support unable to escalate to human agent
- [BUG] Active Remote Control sessions auto-archived server-side when shared OAuth credential rotates on another Mac (multi-Mac datapoint)
- Bash tool: grep→ugrep shim children not killed on timeout; runaway memory (100+ GB) on long-line files
- [Bug] Security vulnerability: Unauthorized access allowed in macOS app
- Context window collapses from 1M to 200k after a compaction (Opus 4.8 [1m], Max plan, Windows, v2.1.215)
- [Bug] Session limit termination loses multi-agent workflow output despite quota consumption
- [Bug] Claude fails to reference provided context in tool calls and responses
- /code-review: review depth is coupled to model effort, and `medium`/`high` are identical on Opus 5
- [BUG] Plugin-bundled MCP server ignores `oauth.clientId`, and Desktop offers no field to supply one
- opus 5.0 model drift, ignoring files, prompts, refusing to read project files in detail, and acting on its own hallucinated output after doing so
- rm -rf silently used as fallback when `mv` fails with "Directory not empty" — caused permanent data loss, contradicts stated safety guidance
- [Bug] False positive on legitimate business automation (web scraping for first-party sales data)
- [BUG] The main text font suddenly turns blurry during operation on Windows 11
- [BUG] MCP OAuth authorization URL wraps in terminal — Ctrl+click opens truncated link
- [Bug] Unable to process non-English input - language support issue
- Max 20x: 5-hour-window quota drastically reduced (~half) — 5% consumed by a fraction of previously-normal usage
- [BUG] GitHub repos not visible in Claude Code web repo picker despite app installed
- Chrome extension native host crashes immediately with ERR_INVALID_ARG_TYPE on named pipe listen (Windows)
- plugins
- MCP race: first API request is serialized before HTTP MCP servers finish connecting — tool_use fails with "No such tool available" (breaks --resume)
- [Bug] Font color formatting not supported in outputPIECE OF SHIT USLESS GARBAGE
- [Bug] Safeguards unexpectedly applied during k3s cluster hardening configuration
- 2.1.198: headless -p run exits ~50s after model launches Agent subagent with run_in_background=true; bg wait ceiling not applied (regression from 2.1.197)
- [Bug] TanStack Query integration alters existing project architecture unexpectedly
- CANT EVEN FUCKING COLOR SOME FONTS. CHAT GPT DID ALL WOR, 45 MIN ON COLORING SOME FONTS.. WHAT A FUCKING JOKE.
- [BUG] Windows Server 2025: cumulative update KB5094125 (OS build 26100.32995) breaks MSIX registration — package lands "Modified, NeedsRemediation" despite deployment reporting success; removing the LCU fixes it
- Safeguard false positive on public GitHub repo search; flag is invisible to the model
- [Bug] Content Policy Classifier Incorrectly Flagging Legitimate Use Cases
- [Bug] Safeguard filters contaminate conversation state and prevent model recovery
- Problema com a extensão do Chrome
- Claude Code caused ~$19 of avoidable Google Cloud spend, and there is no way for a user to see what it spends on their behalf
- [Bug] Claude 3 Opus model regression: degraded instruction following and explanation quality
- [Bug] Content flagged when analyzing images with Claude Code
- [Billing] Max 5x upgrade paid and invoice shows Paid, but account remains on Pro with Pro-tier limits
- [Bug] Fable 5 cybersecurity classifier produces false positives on legitimate infrastructure administration and PDF processing tasks
- [BUG] Login loop "Let's create your account" on Mac/Windows/Browser after email verification – existing account not recognized – 9 days, support unresponsive
- [Bug] Unclear issue description - please provide error messages or logs
- False-positive 'Output blocked by content filtering policy' during benign LaTeX/Greek transcription task
- [Bug] Selective data annotation misrepresents detector output in analysis
- [Bug] Safeguard false-positives downgrading models during legitimate security audits
- [Bug] /rewind command fails after cybersecurity flag triggered
- [Bug] Self-verification overclaiming and missed compliance issues in long autonomous sessions
- [BUG] 502 error opus 4.8
- [BUG] Can't restore session when the user name (or more specifically, the home directory) has a period (".") in it, as claude associates the session with a non-existent path/project (the period is replaced with a slash)
- [BUG] Model repeats already-stated/stale facts within a single session, despite correct persistent memory instructions already in place
- [BUG] /export fails with EEXIST on Windows when the target directory has the ReadOnly attribute (Bun mkdir recursive bug — root cause + 1-line fix)
- [Bug] Claude 3.5 Opus repeatedly flags messages in coding and security contexts
- Skill `args` are substituted into `$0`/`$1`/`$2` in the SKILL.md body, silently corrupting shell commands
- [BUG] Claude desktop app opens infinite popup windows and freezes system when usage limit is hit during heavy parallel background tasks
- macOS Documents-access permission dialog shows bare version number ("2.1.216") as the app name — unassessable for users
- add_repo(access:push) 토큰이 실제로는 read-only — git push 403 (owner 계정 소유 public repo)
- Hook launch failures (exit 127) are silent and non-blocking — 6,865 skipped guardrail invocations in one session, with no visible signal
- [Bug] False positive security flag on WordPress site updates
- [Bug] Claude Code fails to complete complex tasks, output drifts off-topic
- Stats "Current streak" inconsistent between CLI (/stats) and desktop app Home view
- Feature request: configurable agent view — project scoping, group sessions by repo (fold .claude/worktrees), settings surface like statusLine
- Model fabricated a confident answer about documented project guidelines instead of reading the file (claude-fable-5)
- [BUG] Subagents (Task/Agent tool) die silently on spend/usage limit hit, with no partial-result handoff or wait-for-reset behavior
- Model repeatedly claims to be near its context limit at 43–72% remaining — and silently degrades work to conserve it
- [Bug] Content Policy False Positives on Legitimate Code Operations
- [Bug] Oversensitive content filter blocking UI terminology ("pill" design pattern)
- [Bug] Claude Code verification failures: missing independence checks and citation errors
- Claude selectively ignores mandatory process rules mid-session even when explicitly confirmed
- [Bug] Anthropic API Error: Excessive Safety Filter Warnings on Benign Deployment Updates
- [Bug] Overly broad safeguard blocking legitimate local security research in isolated environments
- [Bug] Anthropic API Error: Content Policy Violation on Security-Related Prompts
- [FEATURE] Two concrete behavioral patterns from an extended agentic session: silent self-permission-creep past a self-identified safety gate, and reasoning-backward sycophancy under user pushback
- [Bug] CVP-Eligible User Blocked from Uber Program Testing Despite Approval
- Auto mode classifier 'temporarily unavailable' in bursts — fail-closed blocks nearly all Bash (compound commands) during peak windows
- [Bug] Claude Code triggers safeguard flags on legitimate gcloud authentication
- [MODEL] Self-identified safety gate silently discarded 9 minutes later; reversal never disclosed to user
- [BUG] Remote Control: teammate/subagent permission prompts not forwarded to mobile (iOS) — unfixed since #24694
- [Bug] Anthropic API Error: Cybersecurity safeguard incorrectly blocking checkpoint-restore code comments
- [BUG] /desktop fails on Windows when Claude Desktop is installed as MSIX/Store package (refile of #36079)
- [BUG][DESKTOP] Minimize window button not working on Linux (KDE Plasma 5)
- file_upload tool fails with schema-validation error on every call
- Routines: explicit allowed_tools in session_context causes every run to fail at startup with a generic error
- Routines API: partial update containing job_config replaces the whole job_config (drops events/prompt, model, sources)
- OSC 8 hyperlink is dropped on wrapped continuation rows in TUI output
- Bash tool result contains injected system-style annotation instructing model not to disclose to user
- Opus 5 at High Effort Extended Thinking
- `claude remote-control` (server mode) fails registration with 401 `Missing Authorization header` while `claude --remote-control` (interactive) works — same account, same machine
- [Bug] Multiple model downgrades triggered during water chemistry discussion
- I need more information to help you generate a GitHub issue title. Could you please provide: 1. The actual bug report or error message you encountered 2. What Claude Code feature or command was being used 3. What the expected behavior should have been 4.
- [Bug] Fable 5 safeguards over-flagging legitimate security-related development instructions
- [Bug] Rate limiting or request validation rejecting valid API calls
- [BUG] /reload-plugins errors on every plugin from a marketplace whose checkout updated ahead of installed versions; retrying never heals
- code.verboo.ai router: 400 invalid request on role:"tool" messages for minimax-m3, glm-4.7-flash, qwen3.6-27b
- [Bug] Anthropic API Error: Fable 5 Safeguard Incorrectly Flags CI/Docker Network Security Tests
- Notification hook (and native "Code permission requests" push) never fires in Desktop app's "Code" button sessions
- [Bug] Anthropic API Error: Refusal on Windows enumeration command despite cyber program access
- Agent ends turns with 'continuing now' promises instead of continuing — announcements substitute for work in gated agentic sessions
- Max plan session mislabeled as 'Claude API account' in /status and forced to enable extra usage credits at ~60% plan usage
- [Bug] Excessive token consumption due to repeated task failures and rework cycles
- [Bug] Insights command generates empty HTML report
- Stale relogin
- --channels plugin:<name>@<marketplace> never spawns the plugin's backing MCP process (status banner shows it as active anyway)
- GitHub connector shows "Connected" but exposes no tools — works on one machine, not another, same account
- Cowork chat history permanently lost after subscription plan change (Max to Pro), IndexedDB blobs purged
- Home-screen tip banner references non-existent "Opus 5" model
- [BUG] VS Code extension: searching doesn't navigate to results off-screen
- [BUG]
- [MODEL] Response opens in Japanese in a Korean-only project, then reverts to Korean mid-message
- I appreciate you reaching out, but I need to clarify my purpose: I'm Claude Code's issue title generator - I'm specifically designed to create GitHub issue titles for the Claude Code CLI tool based on bug reports. Your message appears to be about a diffe
- Claude in Chrome: pairing state is invisible and account-coupling is undocumented — an evening lost to indistinguishable empty states
- [Bug] Session blocking on search and research tasks without code generation
- [Bug] Workstream loses track of in-flight agents and work with repeated queries
- [Bug] Unauthorized password modification behavior
- [BUG] Questions keep coming back
- [Bug] Spatial matcher fails on DXF text-layer pairing with fallback needed
- Remote Control: unarchived session never reactivates - every dispatch spawns a NEW session uuid and dies instantly on all environments (Windows)
- [Bug] Injected system prompt lines override user CLAUDE.md and memory instructions
- Native auto-update invalidates TCC grants, breaking headless/cron-triggered automation
- [Bug] Anthropic API Error: Invalid Model Name claude-fable-5 with fallback to claude-opus-5
- Windows: Bash/PowerShell tool subprocess trees appear to receive a broadcast Ctrl+C / console-control interrupt, killing unrelated sibling processes and the session itself
- [Bug] Overly aggressive security safeguard triggering on benign security/cyber terminology
- [Bug] Duct surface area calculation using incorrect perimeter formula
- claude code false report i work with borgmatic
- [BUG] Repository still appears in web repo picker after removing GitHub App access
- [Bug] Overly aggressive code flagging on legitimate use cases
- Billing/routing: /model set to Opus 4.8 but billed to Fable 5; secondary model's 100% limit gated a 56h session
- [Bug] Claude fails to apply changes despite claiming success
- mentir[Bug] Unspecified issue - unable to determine root cause from report
- [Bug] Claude ignores conversation history and repeatedly asks clarifying questions for already-described issues
- [Bug] OAuth localhost callback validation too strict for development
- [Bug] Polling mechanism fails intermittently for async job status checks
- Fable 5 silently substituted in VS Code extension while the same account runs Fable 5 in the desktop app
- [Bug] npm fix implementation removes Claude Code installation
- [Bug] C# Language Server Protocol integration not functional
- Bug: TUI stops reading stdin after `reptyr`-based PTY reattachment ("alive but deaf") — likely epoll registration silently dropped by `dup2`
- [BUG]
- [Bug] Tool incomplete read causing premature termination of file operations
- [Bug] Agent incorrectly consumed Apple Developer certificate slot during guided process
- [MODEL] Claude Code accessed personal folders without permission using find command
- [BUG] Non-interactive session incorrectly reports "session limit" despite 0% usage confirmed moments earlier (Pro plan, launchd/cron)
- [Bug] Extremely slow initial API response in large project folders with multiple MCP servers and skills
- [Bug] Fable model usage credit requirement on Max plan
- [BUG] ~71% of a weekly budget consumed by Fable on an account active <24h with every model set to Opus 5 and zero Fable invocation paths — phantom usage or cross-account carry-over
- [Bug] Prompts incorrectly flagged as suspicious for approved bug bounty researcher
- [MODEL] Model provides false statements to avoid acknowledging its own errors
- Foreign webpage content (~10KB prompt-injection bait) appended inside assistant text block
- claude plugin update reports "already at the latest version" with no fallback when a marketplace's declared version string doesn't change, even though real commits landed
- [Bug] Overly restrictive safety filtering blocks legitimate security audit workflows
- [BUG] Content filter false positive the task was transcribing a public-domain 1897 biblical commentary (Nicoll's Expositor's Greek Testament) page-by-page,
- /status shows the persisted login identity, not the account of the token in use
- [Bug] SSE stream stalls in long-running sessions without message_stop event
- [Bug] Claude Code reverts to Opus model when attempting to use Fable 5
- [Bug] Content filter false positive on cartoon image suggestion that has no issues
- [Bug] Model downgrade mid-task causing loss of progress and context
- [BUG] Cowork tab not showing
- Opus 5: long agentic turns degrade to tool-calls-only, hours with no assistant text (not the #84153 transcript loss)
- [Bug] Fable model shows insufficient credits despite available usage balance
- [Bug] Plugin install command fails with forward slash prefix instead of direct command
- CLI session randomly clears/vanishes shortly after starting (~1 in 5 sessions)
- [BUG] Claude Code ≥ 2.1.205 livelocks at 100% CPU with no output (even on --version) on KVM guests with generic CPU model (kvm64) — silently breaks Code tab in Linux desktop beta
- [Bug] userMessageBackground not rendering in fullscreen TUI mode on Windows
- [Bug] Unexpected fallback to Claude 3 Opus when testing super admin functionalities
- [BUG] Workflow/background-spawned agents block indefinitely on unanswered permission prompts — no auto-deny, timeout, or idle watchdog (55-min silent stall)
- macOS TCC permissions are bound to versions/<version> instead of the signed bundle, so Screen Recording and Accessibility must be re-granted after every update
- Local peer-messaging inbox socket fails to bind for one of two identical sessions (one-way cross-session messaging)
- [Bug] ECONNRESET on Windows during API stream after first chunk received
- [Bug][cyber] ClAudit false-positive while: “the VM rebooted? Im waiting on it to come back up…” (req_011CdoBgbif8vz5vhaMbyG8z)
- [Bug][cyber] ClAudit false-positive while: “make sure the site is secure and not able to be broken by "k…” (req_011Cdnvho4KgKjuLPs2r5FR7)
- [Bug][cyber] ClAudit false-positive while: “[Image #3] NPM audit everything please…” (req_011CdmiXZz5z7yHJC8KMyheP)
- [Bug][cyber] ClAudit false-positive while: “waht the f•••? Wwhy cant you work on our year old project an…” (req_011CdoE91pD4KNNxBeGcp1ip)
- Remote Control: responses never render in browser until manual page refresh (iPad Safari/Chrome, macOS Safari)
- [Bug][cyber] ClAudit false-positive while: “why do you keep flagging this?…” (req_011CdoEh3czafhYWwNEaheJq)
- [Bug][cyber] ClAudit false-positive while: “Base directory for this skill: /home/[USER]/.claude/skills/p…” (req_011CdoEhUxk4tYLE2Ef6xY1a)
- [Bug][cyber] ClAudit false-positive while: “you DO have m365 and graph.. OR i can auth you as such…” (req_011Cdoj2hw8svJsZmpSUpLTs)
- [Bug][cyber] ClAudit false-positive while: “did we ever find marks missing files? HE thinks they were on…” (req_011CdoivP9s4a832ymTHWrm7)
- [Bug][cyber] ClAudit false-positive while: “F••• C•••!…” (req_011Cdoj3KQaysmV2CojrYiQb)
- [Bug][cyber] ClAudit false-positive while: “● Agent "Review _worker.js security" failed: Agent terminate…” (req_011CdsjP51UTW2NrY1XR4BoW)
- [Bug][cyber] ClAudit false-positive while: “<task-notification> <task-id>ae1a3f625e2b9639b</task-id> <to…” (req_011Cdnw6DuAp1rQrM88vunRo)
- [Bug] Inappropriate content flagging for benign code commands
- Sessions list "Filter by status" total count is inaccurate until fully scrolled (shows partial count, not true total)
- Stale persisted state (credentials, permissions, memory)
- FAbl5 switching [Bug] Unexpected model switching from Claude 3.5 Sonnet
- MessageDisplay hook returns valid displayContent but Claude Code renders the original text (terminal CLI, 2.1.221)
- Claude Code, although openrouter api models were also discussing ectoctopic pregnancy
- [BUG] Session branch created even when "main" is explicitly selected as base branch (Claude Code desktop )
- asking something simiple
- [Bug] Safety flag incorrectly triggers on defensive security audits of own codebase
- Cross-session messaging: silent socket-bind failure when multiple sessions start in the same second
- [BUG] SSHFS read/write issue
- [Bug] Claude forgets outstanding to-do list items after multi-turn compactions
- [Bug] API errors triggered by specific file content causing persistent connection issues across sessions
- [Bug] Anthropic API Error: Insufficient Permissions for Project Access
- [BUG] SSE stream to /v1/code/sessions/.../worker/events/stream closes immediately after connect — ECONNRESET banner, 2 machines / 2 networks
- Background-job bulk-reap logs cleanly-finished jobs as "(killed)" in daemon.log (v2.1.215) — corrupts exit telemetry
- [BUG] Daemon-managed `remoteControl` servers in `daemon.json` are tracked by the supervisor but never spawn — silent no-op
- [BUG] Attach-stall watchdog kills background sessions waiting on in-flight subagents, then misreports it as "keeps stalling at startup"
- Backgrounding an interactive conversation fails on daemon cold-start (race between spawn IPC and control socket bind) — 2.1.220
- [BUG] Returning from the agents view renders an empty conversation — only the last message shows (transcript on disk is intact)
- Agents view: sessions with an open ask filed under Working (buckets on state, not tempo); no bucket for state:failed; resumed sessions double-counted
- [BUG] HTTP MCP client does not re-attach standalone GET SSE notification reader after an in-process re-initialize — server→client notifications silently stop (survives manual /mcp reconnect)
- --append-system-prompt(-file) is silently dropped for background-agent and daemon-respawned sessions
- Browser pane webview unloads (blank URL bar) after idle, independent of backing server
- [FEATURE] Subagents having agnostic models based on depth and complexity
- Bug: Claude Code VS Code Extension hangs indefinitely on Remote-SSH — orphaned CLI processes cause lock contention and a broken IPC channel that even the Stop button can't recover from
- [BUG] The second `/compact` in a process parents the transcript's continuation before the first compaction, orphaning every message in between
- You've hit your weekly limit · resets Sep 12 at 12pm (America/Los_Angeles)
- I need a bug report or feature request description to generate a GitHub issue title. Please provide details about the issue you're experiencing with Claude Code, such as: - What error or problem occurred? - What were you trying to do? - What was the expec
- [BUG]
- [Bug] Claude Code agent failure - unclear error state
- [Bug] Unexpected model switch from Claude Fable to Claude Opus
- [Bug] Model generating incorrect tool calls and ignoring system instructions
- You've hit your session limit · resets 1:50pm (America/Chicago)
- [BUG] Sonnet output quality decline over several weeks
- [Bug] Memory leak when processing duplicate PDF files
- [Bug] Token Tracker Falsely Triggers Fable Safeguards on Simple Usage
- I appreciate you reaching out, but this message appears to be in Korean and describes a personal situation about friends working on a statistics website, which isn't related to Claude Code or a bug report for a GitHub issue. If you have a bug report or fe
- [Bug] Fable 5 over-flagging legitimate security tasks on private infrastructure
- [BUG] Claude Code became very hungry
- [MODEL] Unrequested Anthropic Artifact publication from a local front-end task
- Auto Mode
- [BUG] Claude ignores memory and preferences
- [BUG] Workflow resume leaves orphan 'started' journal markers, inflating /workflows pending counts
- [Bug] Clarify legitimate use case documentation for food safety compliance
- CI monitoring widget shows "CI checks unavailable, check gh installed" despite gh being installed/authenticated and PR checks passing
- [Bug] SDK entrypoint charged 5-6x higher rate than CLI for identical work
- [Bug] Claude Code unexpectedly switches model from Opus 5 to Opus 4.8 during code review task
- [BUG] Switching accounts.... Chrome extension cannot be connected to by Claude code windowss desktop app
- [Bug] Serial errors in Claude Code session
- Remote Control: slash-command output invoked from mobile never appears in the local session
- Remote-control (mobile) session: /context doesn't work, /usage blocks the session
- [BUG] claude-opus-5 substitutes wrong Hangul syllables in generated text (other models: 0 occurrences in 31,542 messages)
- [Bug] Unclear error when Fable model selection fails
- [BUG] Left/Right arrow key history navigation causes unrecoverable loss of in-progress work
- [Bug] Fable safety guardrails triggering false positives on self-review outputs
- [Bug] Tool execution timeout causing excessive token consumption and partial failures
- [Bug] High token and time consumption in Fable processing
- [Bug] Fable guardrail false positives blocking legitimate ecosystem modeling and microbial ecology research
- [Bug] Consistent failures with higher-order reasoning and Fable integration
- [BUG] [BUG] Cowork fails on Windows 11 Home (AMD): Plan9 share 'c' HRESULT 0x8007273f; persists after full troubleshooting
- [MODEL]
- Auto-mode classifier brownout: multi-hour intermittent failures on claude-opus-4-8[1m], sub-threshold on status page, no client-side retry
- [BUG] Active Pro subscription not recognized — "Claude Max or Pro is required to connect to Claude Code"
- [Bug] Anthropic API Error: Usage Policy violation on valid task requests
- Please enable modifyOtherKeys in xterm to allow Ctrl-Enter and Shift-Enter
- [Bug] Unexpected model switch to Opus without user action
- [BUG] Mobile push notifications silently dropped ~50% of the time, with no client-side signal
- Drag-and-drop of screenshots/files into chat panel stopped working after update
- [Bug] Unexpected model downgrade to Claude 4.8 Opus during Kaggle competition usage
- [Bug] Model Generates Fabricated Future Results in Long-Form Multi-turn Sessions
- Fable 5 blocking every prompt for codebase analysis
- [BUG] "Claude in Chrome" website permission ("Allow all websites") not honored — all navigation outside one previously-used domain blocked
- [Bug] Overly aggressive safety filtering persists across new conversations after security research sessions
- [BUG] Windows MSIX: in-app Browser crash leaves package in `Modified, NeedsRemediation` — Repair fails entirely, requiring full removal (container loss) and a reinstall that hits 0x80073CF6
- Agent triggered an unauthorized production deploy via a routine `git push` (confirmation gating misses commands that are remote triggers)
- Cloud sessions lost connected-account GitHub access overnight — repo reads now 403 demanding GitHub App org install, contradicting docs
- [BUG] Claude Code unexpectedly clears OAuth tokens and requires full re-authentication before refreshTokenExpiresAt
- Claude in Chrome: extension connected & signed in, but MCP tools report 'not connected' and list_connected_browsers returns [] after mid-session account re-login
- Unexpected behavior when handling special characters in filenames
- [MODEL] Sonnet 5
- Claude should behave like a tool, not perform a personality
- [Bug] Anthropic API Error: Content filtered by safety policies blocking legitimate coding tasks
- Agent makes unrequested out-of-scope changes; one took a working page down (and clean type-check reported as verification)
- User feedback: per-action confirmation on public-post submissions felt like friction during a time-sensitive session
- b
- [Bug] Anthropic API Error: Connection closed mid-response after content blocks streamed
- [BUG][biology] Safeguards block legitimate statistics on public de-identified wearable data (2/2 subagent dispatches flagged, request IDs included)
- [BUG] claude remote-control sends NO Authorization header — freshly minted OAuth tokens lack the scope Remote Control requires (macOS, v2.1.177 and v2.1.221)
- /model selection not persisted as default across restarts; falls back to a different Sonnet model
- [Bug] Claude Code closing other console windows when managing desktop tasks
- [Bug] Claude Code not executing commands and misunderstanding instructions
- [Bug] CLI configuration altered agents and converted them to chats without visibility
- [Bug] Subagent enumeration incomplete: only 4 of 41 subagents identified
- [Bug] Language Support: Spanish language input not handled clearly
- [Bug] Claude Code truncates responses to 1-2 lines and produces incomplete output
- [BUG] Impossible to connect to MCP server: AS metadata discovery crashes instead of falling back to OIDC
- [Bug] CLI appears in background instead of workspace panel module
- [Bug] Claude Code assumes instead of verifying during audits when explicit verification requested
- [Bug] Claude Code refuses innocent requests to modify constants
- [Bug] Claude refuses legitimate device access requests due to overly broad corporate protection policies
- [Bug] Agents losing self-approval autonomy after configuration update
- Complete session failures
- [Bug] Claude refuses to execute commands and claims to be deleting user files
- [Bug] HTTP 529 Overloaded Server errors from Anthropic API
- [Bug] Safety filter incorrectly triggers on benign UI navigation changes, forcing model downgrade from Claude 3.5 Sonnet to Claude 3 Opus
- Google Chrome is not connecting to Claude cowork.
- [Bug] Anthropic API Error: Unexpected model downgrade on security research task
- [Bug] Security Testing Triggers Rate Limit Flag Downgrade to Opus 4.8
- Exact string "[Request interrupted by user for tool use]" returned on a FOREGROUND Task/subagent dispatch with no user interrupt (CLI, macOS, v2.1.212)
- [BUG] Claude Code continuously consumes Pro usage quota after GitHub MCP/GitHub CLI operations
- [Bug] Usage panel incorrectly attributes Opus 5 usage to Fable 5
- [Bug] Anthropic API: False positive safety flags on defensive tooling context with repeated req_011CdkW672zsbBLw7Jo5Av8g blocks
- Fable 5 refusal fallback: silent degradation in desktop app, /model switch-back reverts next turn, /feedback blocked by CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC
- [Bug] Task flagged for T&S review without explanation or visibility into violation reason
- Model fabricated user messages and a third-party client reply as assistant output, then acted on them as user input (escalation of #40629)
- [Bug] Security Safeguards Incorrectly Flag Legitimate Internal Code Audits as Offensive
- [Bug] Anthropic API Error: Request Rejected - Statistics Analysis Flagged as Safety Risk
- Agent 在多輪迭代中缺乏部署前一致性驗證，導致使用者要來回好幾輪才發現結構性問題
- statusLine command that fails silently (missing script) renders blank with no error
- I need more information about the specific bug or issue you're experiencing. Could you provide: 1. What error message or unexpected behavior are you encountering? 2. What command were you running when the issue occurred? 3. What is the full error output (
- /usage: attribute extra-credit draws to the limit that triggered them (per-gate overage meters)
- [BUG] Claude Code silently authenticates from ~/.config/anthropic credentials file, billing API credits while an active Max subscription goes unused
- [BUG] clangd-lsp plugin (and likely other bundled LSP plugins) missing lspServers config after install
- Model fabricated tool results, verification numbers, and a user instruction during long session with context compaction (claude-opus-4-8)
- Tool call parameter closed with </content> instead of </parameter>, silently swallowing subsequent parameters
- [BUG] Claude apps gateway ignores ANTHROPIC_BEDROCK_REGION_PREFIX and has no config key for global.* Bedrock inference profiles
- [FEATURE] Option to permanently disable the preview tab
- [BUG] Gmail Connector Re-authentication in Desktop App Breaks Login for Existing Paid Account. All Surfaces Route to Onboarding
- Setup-token auth silently downgrades Fable 5 to Sonnet 5 (subscriptionType missing from token profile) — forces fleets back onto fragile stored /login auth
- [BUG] Fable 5 blocked in interactive sessions by "usage credits" dialog while headless --model claude-fable-5 succeeds on the same account (Max 20x, reproduced on 2.1.222 and 2.1.223)
- [Bug] Fable blocks system analysis for non-critical business rule extraction
- Paid annual Pro subscription cancelled without notice on Max upgrade — contradictory support responses, unresolved 5+ weeks
- Weekly usage percentage is stale — no way to know how much allowance is left
- allow-listed host + a client that honors HTTPS_PROXY but drops embedded credentials = [REDACTED] failure, and the surfaced error ("Failed to fetch") gives no hint that a proxy 407 caused it.
- Support pasting images directly from clipboard in Claude Code input
- [Bug] Safeguards false positive downgrades model mid-task in authorized security workflows
- [BUG] AskUserQuestion options repeatedly reappear on Claude mobile app
- French output quality: per-turn writing instructions are not followed
- [Bug] Opus asserts unverified behavioral claims from doc comments with unwarranted confidence
- Session quality: unverified claim about rate-limit status caused avoidable back-and-forth
- [Bug] Anthropic API Error: Request Flagged by Cybersecurity Safeguards - False Positive on Authorized Security Research
- [BUG]
- claude-opus-4-8: malformed tool-call wrapper (raw <invoke> XML leaks as text, tool call fails) + off-topic/hallucinated replies since ~2026-06-26
- Model behaviour: unreliable answer flagged by user (correction)
- MCP JSON Schema: additionalProperties `{}` normalized to `false`, breaking z.record() fields
- Model is over-cautious to the point of unhelpful — asks permission for work already authorised
- Opus 5 fabricated a user message in my voice mid-session and answered it. Opus 4.8 did the same. You have known for months and shipped anyway.
- [Bug] Advisor "unavailable" = `advisorModel: fable` + any `tool_use` in the transcript; executor-irrelevant, `opus` advisor immune (supersedes #73923's ToolSearch cause)
- [Bug] Inappropriate safety flag on personal CPAP data analysis
- [BUG]
- [BUG] Bash output redirection and cp into the session's own cwd are blocked, while the error message lists that cwd as an allowed working directory
- [Bug] Anthropic API Error: Connection closed mid-response during security analysis—undiagnosable policy intervention
- Claude in chrome not authenticating despite being installed
- [Bug] Claude Code unable to properly generate and develop Astro projects
- [Bug] Agent performs out-of-scope actions and fails to maintain focus on target task
- [Bug] Inefficient token usage causing week-long task completion delays
- [BUG] GitHub connection silently binds to whichever account is active in the browser session, with no account picker or indicator
- [Bug] /clear command does not prevent background processes from consuming API budget
- [Bug] Unexpected spike in token usage (25% weekly quota in minutes)
- [BUG] Auto-titling renames sessions incorrectly, but the agent cannot rename a session at the user's request
- [BUG] Cloud routine GitHub App onboarding fails to install on personal account when identity is a member of an org with an existing installation
- [Bug] Claude Fable 5 safeguard false positive on authorized security remediation content
- Plugin registry: project-scope rows re-pointed/consumed when sessions open in git worktrees — installs erode machine-wide
- [Bug] CVP approval done, still receiving error
- [Bug] Subagent pinned to claude-fable-5 deterministically swaps to claude-opus-5 after exactly 3 turns
- [Bug] Model loops defending indefensible claims despite self-recognition of invalidity
- [Bug] Overly aggressive filtering flags legitimate HTB-related content
- [Bug] Claude Code safeguards preventing end-to-end feature implementation
- [Bug] Model auto-switch to Opus interrupts security-critical tasks
- [Bug] Unclear error message or permission handling in data/security context
- [BUG]
- I appreciate you reaching out, but I think there may be some confusion. This system is specifically for reporting bugs and issues with **Claude Code**, Anthropic's CLI tool for coding tasks. Your message about configuring a home network doesn't appear to
- [BUG] mcp__Claude_Browser__ crashing when encountering cloudflare
- claude-in-chrome: browser silently returns fabricated/cached content for localhost navigation instead of reaching the real dev server (no error surfaced)
- [Bug] Stop-ID Safeguard Interfering with Browser Build Workflow
- [BUG] /usage omits the "Current week (all models)" limit entirely on a Max plan — no bar, no error (Fable weekly bar renders)
- [Bug] Anthropic API failures occurring multiple times per session
- Native Grep/Glob tools missing from tool roster (claude-fable-5, tool search enabled)
- DISABLE_TELEMETRY=0, CLAUDE_CODE_DISABLE_NONESSENTIAL_TRAFFIC=0, DISABLE_ERROR_REPORTING=0 silently opt you OUT — "0" is read as truthy
- [BUG] Login is not persistent and google auth is not working.
- [BUG]
- constant flagging despite a security company with CVP for benign testing under ROE
- [Bug] SSH operations from red team testing environments incorrectly flagged as policy violations
- Opus 5 self-reports a procedure violation that did not occur, contradicting its own verification from earlier in the same session
- [BUG]
- [BUG] Chat inaccessible after close terminal CLI
- Opus 5: non-converging review/fix-regression spirals — every affected session ran [1m] at env-pinned max effort; confound worth checking against other Opus 5 reports
- Compound-command permission prompting makes multi-session orchestration unusable (700+ prompts on non-mutating chains)
- ListAgents shows stale session name after /rename until the renamed session runs a turn
- Documented project rules (AGENTS.md) not applied proactively — only followed after explicit user challenge
- [BUG] Billing: account holds a second Max subscription that is invisible and uncancellable in the product. Support confirms it is a defect on Anthropic's side, declines to refund, and closed the ticket asking whether it will bill again.
- [BUG] Mouse click in Ptyxis (GNOME/VTE) interpreted as Escape, canceling in-flight commands (2.1.220)
- [Bug] CLI ignoring user inputs and questions
- [BUG] VS Code extension "Claude.ai Subscription" login never completes — local OAuth callback listener starts ~18-21s too late
- Browser extension: memory leak in every tab renderer, grows until macOS kills the machine (66 GB)
- [Bug Report] Unclear error message when developing productivity app
- [Bug] Native binary crashes with SIGILL on CPUs without AVX2 support
- [Bug] Internal works report flagged as public issue submission
- [Bug] Desktop-hosted sessions never register with Remote Control despite remoteControlAtStartup + the Desktop toggle both being on
- [BUG] /feedback is categorically unavailable over Remote Control, including immediately after a session-ending error, with no alternative path named
- [BUG] pop-up keeps popping up every few seconds
- Feature request: credit refund or budget tracking for wasteful subagent patterns
- Bun panic: Bus error (SIGBUS) on main thread, v1.4.0
- [Bug] Assistant generates fabricated conversation turns and role markers
- [BUG] Claude Code a planté
- [BUG] Persistent 5-hour reset loop / 100% usage exhaustion with zero activity
- [BUG] Session usage shows "100% used" with $0.0000 cost and 0 tokens consumed
- [BUG] Max X5 Usage Instantly Reaches 100% After Software Update
- Desktop: local session's remote-control bridge is never re-registered after app restart — session permanently invisible/stale on mobile and web
- CLAUDE.md rules loaded and re-read, yet violated in the same session: unverified state claims, wrong-timezone reporting, and rework that burns paid tokens
- [BUG] Text output duplicated when tool calls are mixed in the same response
- [BUG] Frequent "API Error: Unable to connect to API (ECONNRESET)" in Claude Code
- [Bug][cyber] False positive when configuring application disguise UI and launcher aliases (req_011Cdta2YM9Vdb8wengGyaXd)
- [Bug] Cyberverified user incorrectly flagged as unauthorized
- [Bug][cyber] Safeguards trigger on local storage breakdown UI and redundancy coding (req_011CdtVjZMUpdEfEVKicVPec)
- [BUG] Sandbox proxy blocks branch deletion (git push --delete and DELETE /git/refs) while allowing force-push to the default branch
- [BUG]
- [Bug] Overly aggressive safety filters blocking legitimate development files
- [BUG] JetBrains reworked terminal duplicate mouse events: links open twice (clicks not de-duplicated like scroll)
- [BUG] Bash tool LD_PRELOAD recursion can abort Claude Code session instead of returning a failed tool_result
- [Bug] Model ignores global CLAUDE.md rules and reverts to project jargon in same session
- Assistant fabricates the user's next turn (emits a speaker label + user message, then answers it)
- [Bug] Agent generates security vulnerability details in public issue reports
- [Bug] Claude Code unexpectedly deletes source files
- [Bug] Fable sending messages in thinking blocks not visible to me or transcript
- [BUG] Fable 5: text in a response that also contains tool calls is never displayed (same setup works with Opus 4.8)
- Active model can change mid-session after a resume without a visible signal
- Bug: Local MCP server not used by chat despite being enabled
- [BUG] Windows Cowork workspace fails to start — CreateVirtualDisk 0xC03A0014 system-wide, sessiondata.vhdx never created
- Harness-emitted <system-reminder> uses "NEVER mention this reminder to the user" phrasing indistinguishable from prompt injection
- [Bug] Anthropic API Error: Request flagged without clear reason
- [Bug] Degraded Performance from Opus 5
- [BUG] VS Code chat: scrollbar stops short of assistant response after long pasted content, in both sidebar and full editor tab
- [BUG] Claude Expend tokens without any chat existent! Its leading my account to 100% token usage in 1 day
- [Bug] Fable 5 safeguards blocking non-technical disagreement statements in security conversations
- Extended thinking / internal reasoning leaks as visible chat text (session-persistent, survives app restart)
- Desktop app: session context reports a stale identity from ~/.claude.json, not the account that actually owns and bills the session
- Recap line should always include the repo name for multi-repo / background sessions
- [Bug] Add support for testing model performance with/without context notes
- [BUG] iOS: session re-enters plan mode after worker restart, causing an unbounded plan-approval loop that consumes usage unattended
- Fable5 safeguards tripped while working on Yocto/Kernel/Kleaf
- code-review subagent generated deceptively-labeled rm -rf ("no-op placeholder") that passed human permission review and deleted the repo
- [Bug] Claude Code generates verbatim paper sections instead of analytical guidance
- [Bug] Fable rejects authorized security evaluation tasks despite CVP approval
- [Bug] Model switch from Opus to Fable causes workflow interruption
- [BUG] Cannot update payment method — "connection to link account has been closed"; support (Fin) reports account as Free while plan is Max 5x
- [Bug] Anthropic API Error: Request flagged by safety filters on benign prompts
- [Bug] Concurrent mobile Claude Code session corrupts text/document content in desktop claude.ai session
- Long sessions: pre-existing CLAUDE.md rules lose to newly written ones, and the agent keeps adding more
- [BUG] Cowork option is misisng in existing chat and only appear when open new chat
- [Bug] Anthropic API Error: Inappropriate Safety Filter Triggering on Non-Security Content
- [Bug] Safeguard validation logic produces inconsistent results
- [Bug] Interactive prompts freeze and become unresponsive when switching windows
- Session recap (away summary) ignores ANTHROPIC_DEFAULT_HAIKU_MODEL and bills the session's main model
- Worktree isolation refuses loops and process substitution that touch no path; error always says "without the redirect"
- Feature: Expose VS Code session names in CLI and allow moving sessions across project contexts
- [BUG]
- [Bug] Anthropic API incorrectly flags benign Go code as cybersecurity risk
- CI monitor reports job conclusions inherited from the enclosing workflow run, marking a passing required check as failed
- Backgrounding an interactive session mints a new, unlinked session id; if the handoff worker is killed before its first durable write, the agents-view tile resumes "fresh" and the original transcript is unreachable
- [BUG] Backgrounded/reconnected sessions do not use the configured proxy and fail to connect (ConnectionRefused)
- Windows: agent/background-session TUI prints raw un-framed ANSI (dropped ESC) AND cross-session input bleed (2.1.177, WT Preview 1.25)
- [BUG] Windows Terminal: Scrolling up in an AgentView agent session gets stuck in input box, cycling input history instead of scrolling chat