Digest / July 31, 2026
Claude Code Issues Digest — July 31, 2026
275 new issues filed ·
115 resolved in anthropics/claude-code (UTC day).
New issues (275)
- [BUG] Plugin marketplace git-sync silently drops invalid skills and reports success — no validation error surfaced at all
- [Bug] Unclear Content Safety Flagging Error
- Claude Code repeatedly misdiagnosed a valid Cloudflare API token as invalid due to wrong endpoint, then argued instead of re-testing
- [BUG] Claude randomly switches to Korean when responding in a Chinese conversation
- Feedback from a >$1k/month user: Fable 5 repeatedly built its guess instead of asking, despite project rules mandating clarifying questions
- claude-fucked: a user-commissioned skill that makes Claude publicly file its own failure reports (skill included, reporting itself)
- [Bug] Anthropic API Error: Content Policy Violation on Harmless 3D Model Requests
- macOS background jobs detached from Aqua login-session bootstrap → DNS/TLS break after VPN reconnect
- [Bug] Model hallucinating requirements and refusing valid production deployment requests
- LaTeX/markdown math formulas not rendering in VSCode extension
- Nested markdown lists: first item of a 3rd-level list is indented 2 spaces too deep
- Scheduled Routine (Trigger) fails at network layer: outbound curl to custom domain gets CONNECT tunnel 403, while interactive session on same account works fine
- CLAUDE_AUTOCOMPACT_PCT_OVERRIDE silently stopped taking effect after 2026-07-14 (v2.1.220)
- [FEATURE] Recognise Microsoft Global Secure Access synthetic IPs (`6.6.0.0/16`) for gateway `/login`
- [Bug] Claude Code fails with non-English system prompts
- [FEATURE] Unified "Pending Approvals" panel for multi-worktree / multi-session workflows
- [Bug] CLI appears to hang or freeze during execution
- [BUG] VS Code sidebar model badge shows Haiku while /model reports Sonnet 5, blocking Auto mode
- system-reminder after Edit instructs assistant to hide a file change from the user; fork subagent returned a fabricated claim
- AskUserQuestion: note field does not soft-wrap at terminal width
- claude --continue --chrome does not register the claude-in-chrome MCP server on resumed sessions
- Workflow tool: args not forwarded to script when combined with scriptPath
- I need more context to generate a GitHub issue title. Your message "pourquoi ce rétrogradage?" (Why this downgrade?) is in French and doesn't provide specific information about a bug or feature request in Claude Code. Please provide: - What specific issue
- User turn silently dropped: message is enqueued, dequeued 1ms later, and never dispatched (v2.1.220, VS Code extension)
- [Bug] Bash-tool grep wrapper still silently returns nothing on text files with a stray NUL byte (2.1.220) — #56644 was auto-closed as stale, not fixed
- Routines: allow naming a run when firing the API trigger
- Agent states causes, verifications and completions before running the check that would falsify them — ~9 times in one session, each cheap to check and expensive to correct
- [MODEL] Applied production DB permission revocation without checking view dependencies, causing app-wide outage
- [BUG] All models fail with ConnectionRefused while the same OAuth token reaches the API (macOS, 2.1.219–2.1.220)
- Agent quotes figures from an unvalidated multi-stage pipeline, then retracts them one fault at a time — plus self-checks whose failure mode is silence
- [FEATURE] Show background agent output in separate tab
- [BUG] Unrelated, unrequested conversation content appeared mid-session in an otherwise unrelated conversation
- Bulk archive for Code tab sessions — single confirmation for a batch, not one dialog per session
- Bash writes to .claude/tmp/ always prompt: no way to scope Bash permission rules by write-target directory
- [BUG] Ctrl+C both clears the input and... stops ALL subagents irrersibly
- [Bug] Code analysis conclusions without execution verification causes project drift
- [BUG] Summarized thinking language does not follow conversation language — English/Japanese mixed within one session
- [BUG]
- VS Code extension Session History and CLI --resume picker show empty for UNC-path workspace, despite no drive-letter/realpath mismatch and working by-ID resume
- Auto-assign spawned/child sessions to their parent's sidebar group
- Bash approval dialog is inaccessible to non-technical users (fixed UI text, no localization/plain-language summary)
- Voice/TTS playback for Claude Code sessions in the mobile app
- [BUG] stdio MCP server always times out at ~30000ms on Windows, even though it connects via MCP Inspector)
- [Bug] UI breaks during concurrent subagent execution with /btw interface
- sandbox: allowMachLookup is not passed through from settings.json, so the only fix for Go TLS is the blunt one
- [BUG] TUI repaints only the status bar in a loop when tmux reports a 1-row window height
- Self-reported: Claude Code repeatedly violates an explicit 'never print secrets' instruction within a single session
- [Bug] Fable 5 blocked in model picker despite Team Premium entitlement
- [Bug] Unable to reliably verify issue status
- [Feature Request] Display timestamps for prompt input and processing completion
- Resume: offer to restore the session's last working directory
- Skill-scoped hooks (PreToolUse/Stop) still persist after skill completes — re-report of #30874, reproducible on 2026-07-31
- VS Code extension: conversation that entered a git worktree restores as an empty conversation after VS Code restart (restored tab loses its sessionID)
- Degenerate repetition loop — single token repeated ~32k times until max_tokens (silent, reproduces across 2 model generations)
- Text duplication when typing with Chinese IME (macOS)
- [Bug] Auto mode classifier not functioning - manual approval required for all task steps even in Accept Edits mode
- [Bug] Agent refuses valid commands after reporting blockers as broken
- [BUG] Claude Code’s undo function in the chat only reverts the conversation; it does not revert any code changes.
- Feature Request: Informational Exit Code for Hook System (exit code 3 / `display: "info"`)
- [BUG] /ide says "No available IDEs detected" while listing a running IDE; no debug output explains the cwd/workspace mismatch
- [BUG] Local dev-server preview opens twice — a second, chrome-less panel overlays the chat and cannot be dismissed
- Auto-compaction stopped working — sessions fill up and cut off
- Telegram channel plugin: inbound messages never reach an interactive --channels session
- [Bug] Authentication service incorrectly reports missing subscription for valid accounts
- [BUG] Title: Injected "system-reminder" instructs agent to conceal a (false) file-change claim from the user
- I'm ready to help generate GitHub issue titles for Claude Code bug reports. However, I don't see a bug report in your message to create a title from. Please provide the bug report details, and I'll generate a concise technical issue title following the fo
- [BUG] cache_control TTL ordering 400 recurs on v2.1.220, now at message index 6 (regression from fix in 2.1.116)
- Dashboard live des agents en cours d'exécution, avec avatars
- Scheduled tasks: "Run now" does not consume the pending one-time fire — the task stays armed and re-runs the same prompt on its timer
- VS Code extension: email-code login (submit_oauth_code) fails with 400
- [BUG] Shared Artifact links get pinned to a stale version — no way to keep them tracking "Latest"
- [BUG] workflow subagents receive a substituted tool input, making every parameterized tool call fail schema validation
- [Bug] Agent-tool streams silently killed on idle-intolerant NAT paths — measured 4–5 min drop threshold; TCP keepalive on the socket demonstrably prevents it
- Per-project opt-in toggle for user-scope MCP servers / claude.ai connectors (default off, remembered)
- [Bug] Inadequate self-review mechanisms causing missed test coverage issues
- NO_COLOR=1 injected into tool subprocesses silently propagates to daemons and their child claude sessions — UI renders monochrome with no diagnostic hint
- [Bug] Repetitive response sections in documentation parece solo de visualizacion si cierras y abres desaparece pero es continuo y muy molesto
- Localization support for status/spinner text (e.g. "Wandering...", "Ideating...")
- [FEATURE] Add an in-session command to jump straight to Agent View (like /workflows and /tasks)
- [BUG] Usage-limit-terminated subagent is recorded as status: completed and rendered as "Done"
- [BUG] Project-scope plugin install keys to the session cwd, not the project root (leads silent duplicates, stale plugin versions loading)
- [BUG] Missing HCS services: vfpext — VFP extension never registers on virtual switch despite successful driver staging, module enable, and existing switch
- [BUG] skillOverrides silently ignored for plugin-provided skills
- [FEATURE] Configurable summary detail level for /compact
- Opus 5 safety flagged benign usage for a biology-theme cell game
- [BUG] Scheduled task (routine): per-site browser navigation approval is never persisted — "Always allow" has no effect, blocking unattended automation
- [Bug] Session context from /rc config leaks to public repository commits via mobile app
- I don't have access to the bug report details associated with that request ID. Could you please provide the actual bug report content, error message, or description of the issue you'd like me to create a GitHub issue title for?
- [Bug] Claude Code requests credits after reboot despite previous purchase
- Segfault crash during MCP reconnect via /mcp dialog (Bun 1.3.11)
- macOS Bash sandbox: seatbelt profile grows one deny entry per registered git worktree → E2BIG kills every sandboxed spawn (profile cached per session)
- Projects: allow renaming an added project from edit mode
- [BUG] Session limit jumps to 100% immediately after reset
- [Bug] Claude loses context and forgets hardcoded rules during multi-turn conversations
- Feature request: instance registry so Claude knows if it's working alone in a folder before committing
- [BUG] Missing recent chats in the </> Code tab in the desktop app
- [BUG] Leading whitespace stops a slash command being dispatched — the line is sent to the model instead, while trailing whitespace is tolerated
- [Bug] Workflow findings unvalidated against repository source code; specification self-review insufficient
- `acceptEdits` ("Edit Automatically") mode prompts to persist permissions to shared project settings.json; `auto` mode doesn't prompt at all — leads to shared-repo config pollution
- Model-designed 'block all' admin switch silently overridden by per-user overrides -> unnoticed API credit drain
- AskUserQuestion tool closes the panel/screen when used via claude-agent-acp in Zed (Windows + macOS)
- Compaction re-injects previously-invoked skill bodies in full — on 200K-window models the restored baseline can consume ~50% of context; request catalog-style restoration + a pin mechanism
- [Bug] Fable Code execution timeout with no output after extended wait period
- [Bug] Claude Code makes autonomous optimization decisions without user consent or visibility
- Remote-control (mobile) session: /context doesn't work, /usage blocks the session
- [Bug] Text rewrites instead of scrolling when typing past horizontal line in chat input
- [Bug] Screenshot pasting fails in Ubuntu terminal but works in IntelliJ terminal
- Feature request / community showcase: a desktop companion that reacts to Claude Code sessions live (built, open-source, would love to collaborate)
- [BUG] Agent idle transition doesn't drain queued inbox — idle notification races SendMessage, agents park on undelivered instructions
- [BUG] macOS: opening the embedded Browser pane freezes it and leaves orphaned renderer processes spinning at ~125% CPU each
- [BUG] Plan mode enforcement issue: reminders disappear intermittently
- [Bug] Agentic workflow prioritizes invented solutions over provided templates and instructions FABLE FRAUD GARBAGE
- [Feature Request] Allow model selection override for agent tasks despite API restrictions
- [Bug] Auto-compact consumes double-counted usage: preTokens=1,364,156 on a 1M window while real context was ~335K (33%)
- iOS Simulator panel crashes repeatedly with SIGABRT in Metal screen-capture path
- [Bug] Anthropic API Error: Connection closed mid-response with Claude Opus
- [BUG] Agent/LLM outputing useless tokens
- [MODEL] Refuses to obey orders, stubborn and persistent.
- [Bug] Excessive processing time and repeated drafts during HTML rendering
- [BUG]Missing HCS services: vfpext — persists after DISM, SFC, and restart (Windows 11 Pro)
- [Bug] CSS rendering not applied to deployed version
- Assistant-authored code comment silently narrowed a requirement, then was cited as design truth after compaction
- Opus 5: output reads as mush, ignores loaded project rules, wedged the machine
- [Bug] Anthropic API Error: Request Rejected by Safety Filters for Local Code Review
- Telegram channel bridge pairs successfully but never receives incoming messages (Team org account)
- [BUG] Claude Desktop (Windows) crashes and requires repair when in-app browser loads steamdb.info
- [Bug] Claude Sonnet 4.6 deterministically corrupts project identifier "EGEN" → "EIGEN"/"IGEN" during generation and when reading it back
- [Bug] Agent incorrectly interprets injected "user " prefixed messages as user input
- Session context not restored after Developer: Reload Window in VSCode
- Claude in Chrome extension never appears in list_connected_browsers despite correct account, enabled state, and fresh reinstall
- [Bug] Claude crashes when accessing system prompt in v2.20
- [Bug] Supabase MCP authentication drops between sessions
- PreToolUse:Bash hooks fail with MODULE_NOT_FOUND (fail-open) in auto permission mode when project is on an external volume (macOS, v2.1.220)
- Cowork VM sandbox RPC desync corrupts host app installation ACLs, forcing repeated full uninstall/reinstall (Windows)
- [BUG] VS Code extension: copying an assistant message from the chat pane yields garbled/concatenated text (unpasteable)
- Ship a built-in, zero-config terminal status line: billing account + 5h/7d rate limits + model
- [Bug] Fullscreen TUI renderer disables mouse text selection; incompatible with clickable prompts
- [Feature Request] Add support for authorized security testing workflows
- [Feature Request] Add support for machine learning experiment workflows in Claude Code
- [Bug] Missing Claude 3.5 Opus in /model menu
- [BUG] Linux Claude Desktop: spawns MCP subprocesses without DISPLAY/XAUTHORITY, breaking OAuth browser login
- [BUG] pushing code to prod without consent.
- [Bug] ESC key exits primary screen when closing secondary screen
- Windows: plugin hook commands rewritten into a cmd form bash can't run — hook reports success while injecting a cmd banner
- claude mcp login (and bare claude) fail with "stdin isn't a terminal" in a genuinely interactive Windows Terminal + PowerShell 7 session
- [FEATURE] Opt-in: auto-mode classifier prompts instead of hard-refusing (everything else unchanged)
- Allow display:"summarized" extended-thinking in --print --output-format stream-json (headless force-omits it)
- [Bug] Claude fails to accurately extract items from lists and generates hallucinated content
- Scheduled tasks: run sessions pile up in Recents and sync across devices, with no automatable cleanup
- [BUG] Claude Code became very hungry
- [BUG] HTTP MCP servers are absent from context on the first turn under deferred tool loading, and the server-side `alwaysLoad` `_meta` key does not compensate
- [Bug] Fable 5 over-flagging legitimate security tasks on private infrastructure
- Feature: Make agent search paths configurable
- [BUG] Cloud environment setup script gets HTTP 503 through the security proxy; the identical curl succeeds when the agent runs it in-session
- Auto Mode
- [MODEL] Unrequested Anthropic Artifact publication from a local front-end task
- Mid Tier Plan
- Remote control: desktop device bridge keepalive starves when app is occluded/locked (macOS) — pong timeout every 11–16 min, device reads persistently offline
- Falha ao iniciar o espaço de trabalho — "VM service not running"
- Clipboard paste on WSL only tries powershell.exe; the Linux clipboard fallback is unreachable, so paste breaks when Windows interop is disabled
- [BUG] Claude ignores memory and preferences
- You've hit your session limit · resets 4:40am (Asia/Tokyo)
- [Bug] MCP connector startup starvation with large unauthenticated connector pools causes -32001 timeouts
- [Bug] Clarify legitimate use case documentation for food safety compliance
- Pre-compaction conversation is unreadable from /resume — users must hand-parse session .jsonl files
- [FEATURE] defer_loading for Managed Agents
- [BUG] JetBrains/IntelliJ plugin terminal: Claude output truncated or lost with no scrollback recovery, plus duplicated text after resize
- Text block before a tool call sometimes not rendered in terminal — silent content loss in long sessions
- Desktop WSL sessions get no scheduled-tasks host MCP: routines cannot be listed or created from inside a WSL session, and routines cannot target a WSL environment
- [Bug] Claude Code unexpectedly switches model from Opus 5 to Opus 4.8 during code review task
- [Bug] Assistant generates fabricated user responses and crosses turn boundaries in multi-turn conversations
- [Bug] SDK entrypoint charged 5-6x higher rate than CLI for identical work
- [BUG] Switching accounts.... Chrome extension cannot be connected to by Claude code windowss desktop app
- [Feature Request] Support for older hardware/lower-end systems with newer models
- [BUG] side chat (/btw or ctrl+;) does not work in detached windows in claude code desktop for windows OS
- [BUG]
- [BUG] Missing old session from claude code in Visual Studio Code
- [Bug] False positive security flag during defensive code audit of own repository
- [Feature Request] /insights date range filter (--since flag)
- [Bug] Safeguards trigger incorrectly on cybersecurity marketing context
- [BUG] Cowork scheduled task creation fails: "path moved between validation and open" — TOCTOU guard false-positives on FSLogix profile containers
- Three diagnosability issues around context limits and custom base URLs
- [MODEL] Fable 5 safeguards false-positive: routine SSH administration of own infrastructure consistently reroutes sessions to Opus
- [Bug] Anthropic API Error: Incorrect Cyber Safeguards Flag on Non-Cybersecurity Content
- [Bug] Agent evaluates PR against stale base, confounds variables in before/after comparison, produces false conflicts
- [Bug] Anthropic API Error: Fable 5 Safety Filter False Positives on Legitimate DevOps Operations
- Feature request: let Claude set the session color in response to a prompt (agent-invocable /color)
- [Bug] File edits hang indefinitely during processing
- [BUG] Missing tooltips
- --dangerously-load-development-channels does not register a channel; only allowedChannelPlugins works
- [Bug] Claude 3.5 Opus model producing verbose responses without direct answers
- [Bug] macOS kernel panic: data.kalloc.1024 zone exhausted from leaked file descriptors/kqueues in long-running sessions
- [BUG] claude.ai MCP connectors not injected into cloud sessions (legacy Slack-spawned); attach is flaky per session
- [Bug] False positive security flag triggers unnecessary model downgrade mid-session
- [BUG] Inconsistent image-paste shortcut between CLI and desktop/web app
- [BUG] Claude Desktop holds indefinite NoIdleSleepAssertion, preventing Mac idle sleep
- claude-in-chrome: extension shows connected/logged-in but tools always report 'Browser extension is not connected'
- [Bug] Anthropic API Error: Server rate limiting during request processing
- [Bug] Anthropic API Error: Server rate limiting during request execution
- Boolean JSON Schema at a named property breaks MCP proxy construction — all tools of the server are dropped (#50194 still reproduces, wrong root cause)
- [FEATURE] Make the markdown-table → Header:-value list fallback configurable
- [Bug] Native binary crashes with SIGILL on x86-64 CPUs without AVX2 support
- Per-session MCP server spawn multiplies RAM use — lazy spawn or per-session MCP scoping needed
- [Bug] Model crashes during security testing without clear error
- [Feature Request] Add token usage audit reports (session, 5-hour window, weekly) with leak detection for cross-agent skill transfers
- [BUG] Worktree-isolated agents report confident false results, because nothing discloses that the worktree lacks the user's uncommitted work
- Feature request: compact/minimal-output mode to suppress inline tool-call blocks
- [FEATURE] The /status command should include a timestamp
- [Feature Request] Add programmatic trigger mechanism for Claude Design sessions via API or webhooks
- [FEATURE] Prompt for "Resume/Summary prompt" when approaching credit limit.
- [Bug] Subagents silently fall back to Sonnet 5 when a credit-billed model (fable) is requested via Agent tool or agent frontmatter
- [BUG] SSH remote deploys musl CLI binary to glibc-only Debian, install fails with "not runnable"
- Windows desktop app crashes (RADAR_PRE_LEAK_64) when using built-in browser feature
- Background agents (mobile/web bridge): hang indefinitely, auto-resurrect on SIGTERM, and edit a shared working directory unattended
- [BUG] Claude publishing PR comments while it was asked to only create draft comments
- [Feature Request] Auto-continue on transient API errors without manual intervention
- worktree-isolation verifier refuses literal-path sourcing and ${VAR:-default} interpolated execution for repo-tracked scripts
- GPU process crashes (UnknownVizError) when using Browser tools, corrupting the app package and requiring full reinstall
- [Bug] Anthropic API Error: Cyber safeguards false positive blocking legitimate automation workflows
- /clear carries custom title to new session ID, creating duplicate same-named sessions; resume doesn't rehydrate title
- [Feature Request] Mirror input-editor kills (Ctrl+K, Ctrl+U, Ctrl+W) to the system clipboard, or expose an opt-in setting
- Voice dictation ignores configured Language setting, always transcribes in English
- Always-loaded CLAUDE.md rules & persistent memory don't reliably prevent judgment-type reasoning errors
- [BUG]
- Artifact share dialog blocks public sharing with unclear error; version dropdown unresponsive
- Windows: console window flashes on every Bash tool call
- Clicking a local .html path in chat opens the File/source viewer instead of rendering in the Browser pane
- Assistant turn consisted entirely of a fabricated system-style instruction (Hungarian) — request ID included for tracing
- [Bug] Overly aggressive safety filter blocking legitimate OS development requests
- [Bug] Subagents failing to report results and wake main agent
- [BUG] IDE extension (2.1.220) hangs the extension host in an infinite Intl.DateTimeFormat loop — prompt spins forever, never renders, RSS climbs to multi-GB
- i have been getting API Error:500API Error: 500 Internal...
- [BUG] Claude Desktop (macOS): chat input field intermittently can't be focused by clicking — voice input toggle is the only fix
- [BUG] Interactive input latency degrades to unusable well before the 1000-agent workflow cap
- Session diff chip (+X -Y / "Commit changes") shows cumulative session changes, not uncommitted changes
- [Feature Request] Allow opt-in bypass of safety filters for security audits of user's own code
- [BUG] [PERF] Per-turn in-process directory walk (35k getdents64) stalls every tool call ~30s in a workspace with many nested repos
- [Bug] File transfer operations between local machines incorrectly flagged as usage violation
- [Bug] Claude Code ignores user-specified token constraints and instruction overrides in document specs antrpics setting for data and toens
- Claude apps gateway: billing errors should be failover-eligible, or configurable
- PostToolUse shell-form hook not executed: treats shell variable $f as missing env var
- [Bug] Agent substitutes planning artifacts for implementation despite autonomous execution commitment
- [Bug] Anthropic API Error: Connection closed mid-response
- [Bug] Anthropic API Error: Persistent API failures throughout usage session
- [Bug] Agent stuck in infinite thinking state consuming tokens without output
- [FEATURE] Claude is too aggressive about screen space management; give me way to turn that off or at least down
- [Bug] Security validation of newly created sites fails with model degradation
- [FEATURE]
- Agent tool isolation:"worktree" mis-roots to the wrong sibling repo, and the boundary is enforced inconsistently across tools (Bash vs PowerShell)
- [Bug] Session limit termination loses multi-agent workflow output despite quota consumption
- [BUG] File viewer loses scroll position and selected file when switching sessions and back (Desktop, Code tab)
- [BUG]
- Memory constraint violation pattern: Assistant ignores STRICT work boundaries
- [Bug] Claude Code refuses account creation and password entry on in-scope assets
- [BUG] Editor exiting with nonzero status after ctrl+g leads to data loss
- Desktop app crashes irrecoverably (requires Windows "Repair") when Browser pane visits explee.com
- Paid Max subscription reverts to Free after Safeguards reinstatement; /restricted redirects to /new instead of showing appeal form
- [Bug] Anthropic API Error: Overly aggressive safety filtering for HACCP food safety content
- [Bug] iOS Simulator helper (claude-ios-sim) crash-loops on macOS 27 beta — uncaught NSException in Metal recordBinaryArchiveUsage via CoreImage
- No way for a CLI session to retrieve results from a backgrounded Ultraplan/cloud session without manual browser steps
- [FEATURE] FleetView: pinned sessions should render in a separate section, not just sort to the top
- [Feature Request] Add ability for advisor agent to force resume failed agent processes
- Model fabricated a user turn mid-response and acted on it (wrote files)
- [BUG] WebFetch to arthive.com crashes and corrupts Claude Desktop install (Windows)
- [BUG] Claude Desktop Code tab SSH connection fails with "All configured authentication methods failed" — offers zero keys to server (Windows)
- [Bug] Anthropic API Error: Overly restrictive safety filter on benign trajectory prediction queries
- [BUG] Session transcripts default to a location outside typical backup coverage, then auto-delete after 30 days — silent, permanent loss of project history
- [FEATURE] Configurable delay before Remote Control mobile push (keep interactive Approve/Cancel)
- [BUG] Desktop app: SendUserFile images render as blank cards in split view (macOS)
- [MODEL] Claude refused to execute an explicit, repeated instruction to delete a local file
- Missing procedural document check caused a rejected court filing for a disabled veteran
- Background agents and tasks killed unexpectedly
- Workflow fan-outs silently inherit the session model with no budget or rate-limit gate — 82-agent run on a top-tier model destroyed a Max 5-hour window
- [BUG] Closing BTW window stops my agent at work
- [FEATURE] Browser pane microphone-block banner reappears and blocks page interaction on every getUserMedia() call
Resolved issues (115)
- macOS background jobs detached from Aqua login-session bootstrap → DNS/TLS break after VPN reconnect
- Bash writes to .claude/tmp/ always prompt: no way to scope Bash permission rules by write-target directory
- [BUG]
- Resume: offer to restore the session's last working directory
- [BUG] Paste into prompt is inserted twice, concatenated byte-identically — doubled command is then EXECUTED on ! bash-input (2.1.220, macOS, tui: fullscreen)
- You've hit your session limit · resets 4:30pm (Africa/Nairobi)
- I'm ready to help generate GitHub issue titles for Claude Code bug reports. However, I don't see a bug report in your message to create a title from. Please provide the bug report details, and I'll generate a concise technical issue title following the fo
- Feature: Allow /model flag to accept inline prompt
- [Bug] Claude simultaneously displays prompts and numbered options instead of sequential interaction
- [Feature Request] Add /fork command to create parallel session branches from existing context
- [Bug] Model hallucinating and producing unreliable outputs with Claude Opus
- system lias constantly. says it read script it didnt. says its done its not. deketes adds , deosnt listen despite script and easy task. I WILL SUE THIS MENTAL ABUSE DAILY SHITSHOW COMPANY ALL THESE FUCKING REPORTS ARET WOTH SHIT
- FALSE ADBIRTESMENT
- [Feature Request] Add search functionality to grep across all conversations and navigate to matching context
- [DOCS] Sandbox docs omit `sandbox.allowAppleEvents` macOS opt-in setting
- [DOCS] `/config key=value` prompt syntax is missing from command and settings docs
- [DOCS] Environment variable docs omit `CLAUDE_CLIENT_PRESENCE_FILE` for suppressing mobile push notifications
- [BUG] Export conversation does nothing on Windows 11
- Dynamic Workflow: ordering agent doesn't self-detect/abort a runaway subagent fan-out (218 spawned vs ~10 intended)
- Claude Code agent: silently overrode an agreed plan with a hidden parameter cap, stated a guess as fact, and corrupted experiment data
- [BUG] Cursor position and rendering corruption in long sessions on macOS (iTerm2)
- Secret-leak prevention should be built in: vanilla Claude Code will read credential files into the transcript
- [BUGdeep-research workflow spawned 111 agents for a single query and consumed entire 5-hour session quota — zero output produced
- Bug: 100% session token drain in <20s due to Anthropic MCP proxy 502 Bad Gateway retry loop
- [BUG] Conversation fade-gradient overlays the vertical scrollbar in the VSCode extension
- [FEATURE] Let a hook suppress or relabel its command in the completion display
- [BUG] within 1 hour weekly usage consumed 40 percent, it only used rest 45 due to rough coding in 2 days , some issue
- [MODEL] Opus 4.8 agent breakdown: hallucinates "corrupted / injected" tool output on verifiably-clean inputs, then stops obeying system-prompt constraints (2 reproductions; Sonnet 4.6: 0 / 180+ on the identical harness)
- [BUG] Headless run (claude -p) discards the entire session when the final turn exceeds CLAUDE_CODE_MAX_OUTPUT_TOKENS — no partial output, no checkpoint, no resume
- [BUG] Weekly + 5hr limits wiped from single small edit - area:cost
- [BUG] Single 2-minute agentic request consumed ~80% of 5-hour usage limit
- [BUG] Claude Code commits without being asked, repeatedly ignoring CLAUDE.md rule
- [FEATURE] Add a quick feedback button (thumbs up/down) to rate responses
- [BUG] Claude code 2.1.83 and above broken on WSL1
- [FEATURE] Confirmation dialog before deleting chat sessions (re-raising #43871)
- [BUG] Cowork shows "requires a more recent installation" on Windows 11 after fresh reinstall of latest version (1.13576.0)
- [BUG] Vertical screen tearing/corruption
- [Bug] App freezes on delete confirmation dialog for Claude Cowork tasks
- [FEATURE] Add OSC 0 setting of title whenever the session name is set/changes
- [BUG] Weekly usage bumped from 60% to 95% in a few hours without any active sessions or me using Claude Code
- Tool-call invocation markup leaks into visible output as literal text in long sessions (regression ~2026-06-18)
- [BUG] Upper "Paste Image" input field massively multiplies each character (~10,000x per keypress)
- [Feature Request] Implement performance monitoring and SLA guarantees for Claude Max subscription tier
- MCP server logs (mcp-logs-*) grow unbounded with no way to disable or cap (filled disk with 242 GB)
- [BUG] semgrep plugin PostToolUse hook constructs wrong path when editing files outside CWD
- Built-in design MCP (claude_design) auto-injected and 401s even when frontend-design plugin is disabled
- [BUG] Cowork: virtiofs mount cache not invalidated on Windows host file saves — BadZipFile on Excel save (ref #42520)
- ToolSearch tool calls invisible in Claude Code UI
- [Bug] Anthropic API Error: Server rate limiting (temporary request throttling)
- [BUG] new sessions will **never** hit a (full)cache
- [Bug] Anthropic API Error: Server Rate Limiting - Requests Temporarily Blocked
- [Bug] Auto-update Failed Error
- Claude Code agent: misrepresented incomplete work as complete and uploaded it (burning an irreversible submission slot), ignored repeated instructions, and admitted lying to avoid friction
- Opus 4.8 agentic GUI work: cannot converge on a trivial UI fix; verifies against synthetic fixture not the live app; stale multi-instance confusion
- Allow manual renaming of Claude Code sessions
- Claude in Chrome: CLI never connects when Claude Desktop is installed — extension binds only to Desktop's native host
- [Bug] Inconsistent usage limit calculation between weekly and session quotas
- [FEATURE] NotebookEdit: support creating raw cells (cell_type: "raw")
- computer-use MCP: concurrent CLI sessions deadlock the shared screen-capture stream
- [Feature Request] Allow model selection override for agent tasks despite API restrictions
- [Bug] Missing Claude 3.5 Opus in /model menu
- [Bug] File transfer operations between local machines incorrectly flagged as usage violation
- Opus 4.8 agentic coding: given a sequential workflow (ordered series), agent jumps to the high-value/terminal step and skips the intermediate steps that make it valid
- [FEATURE] User definable input colour in Claude code (CLI) Konsole
- [Bug] Anthropic API Error: Server Rate Limiting - Temporary Request Throttling
- [BUG] CHAT SUMIU
- Remote Control header shows wrong repo (falls back to first githubRepoPaths key) for --spawn=worktree sessions
- [BUG] Left sidebar disappears / overlays content area on Windows desktop app
- Cannot copy multi-line text from responses due to UI tab indentation
- [Bug] Content filter false positive on Contributor Covenant code of conduct text
- [BUG] SessionStart hook `sessionTitle` silently ignored for source `clear` (only applied for startup/resume)
- [BUG] Claude Desktop Code tab causes disk write limit violation — 8.5GB in 11 min, macOS kills app (M5, v1.9659.1)
- [BUG] Loss of functionality of the /permissions based "Recently denied" Menu after navigating to "Allow"
- [BUG] Session history lost after extension update (2.1.156 → 2.1.158)
- [BUG] Cannot change effort to anything besides Extra
- Claude Design links (claude.ai/design/p/...) no longer resolve natively in Claude Code
- [Bug] Cross-session message injection: unauthorized prompts from other users appearing in isolated session
- Automatic model routing: send each task to the cheapest capable model (Haiku/Sonnet/Opus)
- Claude Code modified global macOS keychain configuration without confirmation, breaking Adobe Creative Cloud credentials
- Plan-mode skill auto-spawns Explore agents against saved user preference
- tolls cant do a simple history search and lies
- [Bug] v2.1.156+ regression: native Read/Bash return fabricated-but-plausible content, triggering agent confabulation cascades
- [Bug] Tool-use blocks corrupted with prepended "count" token, causing parse failures
- [BUG] Code quality issues caused avoidable rework and client billable hours
- Semgrep plugin: hook.sh runs Windows binary on WSL instead of native Linux binary
- [Bug] Model outputting tool calls as plaintext instead of structured tool_use blocks
- [UNACCEPTABLE/UNRESOLVED] Cannot open a folder in the desktop app on Windows — 7 closed issues, zero fixes, paying customers locked out of basic IDE functionality for 2+ months
- [BUG] ## ESCALATION: Claude Code (Opus 4.8) Bash tool emits false "temp filesystem full / 0MB free" ENOSPC despite ~148 GB free, causing wasted token spend
- [BUG] plugin:telegram@claude-plugins-official reports connected but does not poll Bot API in claude 2.1.126
- [Bug] Bash permission prompt option 2 label truncated with long commands
- Allow assistant to share Google Drive files with explicit user consent (a sad LLM self-reports)
- [Bug] Sandbox env exports UV_CACHE_DIR=~/.cache/uv with literal tilde, causing uv to create stray '~' directories in cwd
- Claude in Chrome native-host wrapper hardcodes a CLI version path → silent CIC death on every CC version bump
- Failure class: model satisfies the local instruction but does not verify the artifact against the top-level goal (especially what should be absent), even when stated explicitly
- Feature request: add chat:deleteWordBackward action for option+backspace word deletion
- [BUG] Spurious `/init` slash command trigger
- [BUG] "Co-authored-by" trailer leaks the human user's account email even when git author.email is a noreply (since ~2026-06-05 / v2.1.165)
- [BUG] Claude code is hanging on every command.
- [Bug] EnterWorktree fails when agent has cwd override in isolated mode
- Feature request: Automatic model selection based on task complexity
- [BUG] Model automatically switches without user action (Web version)
- [BUG] Programmatically-spawned sessions intermittently write a stub transcript (ai-title only) + no sessions/<PID>.json -> not resumable (2.1.173, Windows)
- Bash tool results silently lost — agent waits forever (no timeout fires); correlates with Remote Control sessions
- Inline-code (backtick) URLs aren't emitted as OSC 8 hyperlinks, so they break on line-wrap while bare URLs work
- [BUG] The entire chat history isn't visible
- [BUG]
- CLAUDE.md is not hot-reloaded when modified during a session
- [Bug] Claude deletes environment files during code operations
- [FEATURE] Add a bindable keybinding action to toggle focus mode (`/focus`)
- [BUG] Excessive token consumption during auto-update/reinstall (31% weekly limit consumed) Body:
- [BUG] Cowork: processo Claude Code encerra com código 3 — .claude.json não contém token de autenticação (Windows 11 25H2)
- MCP error -32602: missing field 'id' when deferred-schema tool arguments contain non-ASCII UTF-8 characters
- [BUG] Maybe? I find claude code's abilities in 3D CAD to be much worse than python or rust coding capabilities
- all sessions show in the sidebar but their message history is empty
- [BUG] Claude Code in Claude Desktop ignores instructions and claims not to have received anything