Digest / July 28, 2026
Claude Code Issues Digest — July 28, 2026
320 new issues filed ·
155 resolved in anthropics/claude-code (UTC day).
New issues (320)
- Feature request: memory should inherit like CLAUDE.md does (global + project layers)
- [Bug] Claude opus 5 generates low-quality code with excessive bugs
- I appreciate you reaching out, but I need to clarify that this doesn't appear to be a bug report for Claude Code, the Anthropic CLI tool. Your message describes an internal phishing simulation campaign at your organization, which isn't related to Claude C
- [BUG] Crash When Auto-Update Runs and Checksums Dont Match
- Windows: worktree creation in an LFS repo creates a literal dev\null\ directory
- [BUG] Auto-generated session name matches an unrelated project file, not the conversation — identical name reused across sessions in the same cwd
- Add native spellcheck support to the chat input box in the VS Code extension
- [BUG] Explicit model pin claude-fable-5[1m]
- [BUG] VS Code Chat Sessions Right Hand Column
- [BUG] Windows desktop: markdown links with absolute-path hrefs are styled clickable but clicks are silently dropped (chat + .md preview); auto-linkified plain paths in the same pane work
- Hooks cannot block quietly: `decision: block` always renders a user-visible notice in the CLI (`suppressOutput` has no effect)
- [BUG] Pasting snips or simple plain text into Claude code chat causes VS code to freeze and requires a restart
- [Claude 5: Opus 5/5.5 + Fable 5/5.1] Continuation of #57902 — receipt-ignoring, verdict-overclaiming, label-laundering, invented-deferral persist
- [BUG] Desktop SSH remote session: NO_INSTALL_RESULT with zero remote filesystem activity (Docker container target, distinct from #77753)
- Marketplace plugin agents silently de-register on session resume when the marketplace is auto-updating; no re-registration path for the rest of the session
- [BUG] Cowork on Windows: mid-session folder grant refuses the home directory that the task-start picker offers
- [BUG] Inconsistent spelling ("judgement" vs "judgment") in generated recommendation text
- [BUG] ralph-loop never loops on 2.1.x: Stop hook reads the transcript before the current turn's assistant message is written, then deletes its own state file
- [BUG] ralph-loop: --max-iterations 08 silently discards the cap and makes the loop unbounded (leading zero parsed as octal)
- [BUG] ralph-loop: a bare message equal to the completion promise ends the loop with no <promise> tag (perl -p prints unmatched input)
- [BUG] ralph-loop: the expected completion promise is not whitespace-normalized, so a promise containing a double space can never match
- [BUG] ralph-loop: iteration counter freezes when the state file reads "iteration:N" (writer stricter than reader), so max_iterations is never reached
- [BUG] Cowork/Code fail with 403 "Invalid authorization" for 10+ days — Chat works fine
- [BUG] Repos not appearing in Claude Code web repo picker — "No repos match" / "일치하는 저장소가 없습니다"
- [BUG] macOS: sleep inhibitor SIGKILLs its caffeinate before spawning the replacement — Mac idle-sleeps mid-task once the display is off
- Auto-memory is inconsistently loaded in git-worktree sessions (same repo, same day)
- permissions.ask is silently skipped for git when the command is prefixed with `cd <other-dir> &&`
- [FEATURE] Let the desktop app surface on-disk transcripts (cross-machine session continuity)
- [BUG] Claude Desktop (Windows) — GPU process crash leaves main process hung, requires 'Repair' to relaunch
- [FEATURE] Register Claude Desktop as a valid target for Windows 11's native search/Copilot key remap
- [BUG] Subagent replies split by the output-token ceiling: only the last assistant message reaches the caller, the rest is dropped silently
- Coding agent repeatedly reports 'process launched' as completion for a drive-to-output task, never verifying the rendered output
- In-app Browser pane crashes Desktop app's GPU process (exitCode 101457950), even with hardware acceleration disabled
- [BUG] Opus 5: a user correction is absorbed as a topic, not executed as a task — and checkable premises are issued to subagents unchecked
- run-skill-generator: step-0 probe cannot see an existing skill in the target unit, silently rewrites it
- Transcript JSONL corruption; records spliced by unsynchronized writers
- [Bug] Code review fails on WebRTC client implementation
- [BUG] Remote Control: messages typed locally while a turn is running are never synced to connected devices
- [Bug] Project memory updates not synced to global memory
- [Bug] API Connection Error: ECONNRESET During Request Retries
- [Bug] API Error: Connection Closed Mid-Response
- Coding agent skips the repo's own governing documents and working examples, reimplementing from memory and repeating the same failure
- [Bug] Claude Code fails to retain instructions across multiple executions
- [BUG] asking for permissions every 3 seconds while in bypass mode
- `tools:` allowlist is enforced for subagents but silently dropped for named agents (teammates)
- [BUG] Fable 5: text in a response that also contains tool calls is never displayed (same setup works with Opus 4.8)
- [BUG] Desktop app (macOS): copying text in the in-app Browser pane silently fails — the clipboard is never written
- [BUG] Assistant fabricates conversation turns in its own output (leaked `user`/`th` role tokens), including a self-generated injection payload
- [FEATURE] Allow individual users to opt out of org-level customized spinner verbs
- [Bug] Cmd+click file paths not opening in fullscreen TUI on Warp (macOS)
- [BUG] Cannot login via Desktop App
- [Bug] OTel claude_code.user_prompt only emitted for slash-command dispatches; regular prompt submissions emit no event
- `-p` + `--dangerously-skip-permissions` still emits interactive-style prompts in output
- [BUG] VS Code extension: Fable 5 blocked behind 'requires usage credits' on Max 5x plan - consent dialog never renders, /model hint is circular
- [BUG] Read deny rule bypassed by IDE selection context when file path is reported WSL-style (/mnt/c/...) on Windows
- [Bug] API Error 400: MCP Tool Reference Not Found in Available Tools
- iOS Simulator panel shows black screen after host reboot, doesn't recover even after underlying simulator is healthy again
- [Bug] Auto Mode Blocks Edit Tool When Safety Classifier Model Unavailable
- [Bug] Claude Code refuses to execute commands and reports system destruction
- [Bug] Project Access Blocked - Unable to Deploy Production Build
- Session that died mid-turn is silently adopted as a background agent (isMeta "Continue from where you left off." auto-injected), locking the original out of /resume — v2.1.220
- [BUG] [BUG] Billing: extra usage charged during acknowledged July 17 config error — compensation issued as unusable credit, refund to payment method requested (EU)
- Auto-restarted session gets stuck on an exhausted default model; 89 min downtime, manual recovery required
- [FEATURE] Claude should accurately be able to respond to all claude billing/licensing request information regardless of the model
- Named Agent dispatches silently discard the subagent's final text, and `run_in_background` is ignored on that path (v2.1.220)
- [Bug] Agent list view drops completed sessions too quickly
- [BUG] Cowork VM service repeatedly tears down on idle; recovery is a slow cold-boot; service DACL blocks any external mitigation
- Windows MSIX: failed auto-update while app is hung uninstalls Claude Desktop (RemoveForAllUsers) - app disappears, daily reinstall required
- [Bug] Cyber safeguards falsely blocking subagents on defensive security work
- [FEATURE] Inline side-panel preview for .docx/.pptx/.xlsx sent via SendUserFile (pipeline already exists in Cowork)
- Cowork sandbox recurring failure — useradd exit 12 / no space left on device (5 recurrences in a week, prior fix didn't hold)
- [BUG] teammatemode in-process in global .claude/settings.json is ignored in psmux session
- Scroll issue PageUp/PageDown don't scroll the conversation, only the input text
- [Bug] Artifacts with mermaid blocks cannot be shared publicly
- [Feature Request] Add UI padding/density customization settings
- [BUG] Claude Code desktop (Code tab sidebar): "Sort by" has no effect on project group ordering — groups are always alphabetical
- [BUG] TUI visual artifacts during diff view in PyCharm integrated terminal
- [BUG] Chrome pluggin doesn't more communicate with Claude code
- Claude in Chrome injects a page-scoped `.hidden { display: none }` rule, breaking Tailwind `hidden md:flex` layouts
- [MODEL] Opus 5 re-reads background task status ~25 times in one turn instead of ending the turn and waiting for the notification
- [Feature Request] Setting to hide Edit/Write diff and content previews in the transcript
- Model performed unrequested destructive fix: asked to 'check' desktop app status, it force-killed the app's processes
- Insufficient contrast for sidebar labels and icons (accessibility)
- [FEATURE] Proposal: Treat Project Instructions as an Always-Active Skill
- [MODEL] sonnet 5 Extra
- [Bug] Code highlighting navigation routes to incorrect file
- [Feature Request] Multi-file multi-line code selection highlighting like Cursor IDE
- AI quoting tool output B2B cost as client price, causing real financial loss (NT$10,000+)
- [FEATURE] Live tokens-per-second (TPS) display during streaming
- False 'auto-update failed' notification when already on latest version
- [BUG] Resume/attach silently overwrites background agent session titles with another session's name
- [Bug] Claude Code error handling for Opus model reliability issues
- [FEATURE] Configurable pageSize for AskUserQuestion tool
- [Bug] Unclear Issue Description - Please Provide Error Message or Log Output
- [BUG] Desktop app: session worktrees ignore `worktree.baseRef: "head"` and branch from origin/HEAD
- SONET AND OPUS ON CLAUDE CODE CREDITS DISSAPEAR ABOUT 100 EURO
- [BUG] Ctrl+click on a hyperlink opens it twice in Windows Terminal (WSL2): Claude Code fires on mouse-down, terminal fires on click completion
- [MODEL] Multiple self-corrections in one response
- [BUG] Desktop app reinstall silently loses all Claude Code session list entries — transcripts intact on disk but unrecoverable from the UI
- Dispatch missing
- Multi-minute silent stalls on claude-fable-5; no client-side indication of waiting
- Fleet/agent sidebar view misrenders and mouse clicks misalign above a window-size threshold (Windows Terminal)
- MCP tool call cannot pass JSON null for a required parameter typed anyOf:[integer,null] — arrives as the string "null"
- Assistant response occasionally includes a fabricated user turn
- Feature request: Chinese (zh-CN) UI localization
- Custom groups/folders for session sidebar (recent items list)
- Feature request: Add Chinese (zh-CN) language option for the UI
- Bundled ugrep 7.5.0 busy-loops at 100% CPU forever when --ignore-files resolves to a directory (e.g. a dir named .gitignore); orphaned children saturate the machine
- [BUG] Resume drops expanded Skill body when Skill is batched with another tool call
- [BUG] Subagents execute independent tool calls sequentially even when explicitly instructed to batch — including fork mode with identical system prompt
- Improve text-selection highlight contrast in Dark Mode
- Document/expose available Markdown formatting keyboard shortcuts
- [Bug] Agent executes large refactors without plan approval despite repository guidelines requiring incremental changes
- Usage/credit limit shown in Claude Code does not match claude.ai (21% used vs. blocked)
- HTTP MCP OAuth reconnect fails with "MCP endpoint not found at <origin>" right after successful token exchange
- [BUG] Misleading `plugin-cache-miss` ("not cached at <existing path> — run /plugin to refresh") when the real cause is a missing install record for the current project
- remote-control --spawn worktree: default pre-created session dies with 404 and leaves a stale .claude/worktrees that permanently breaks all later spawns (EEXIST, Windows)
- [Bug] Missing zeroization on AEAD post-decrypt verification failure
- I appreciate you reaching out, but I need to clarify something important: I'm Claude, an AI assistant made by Anthropic. I'm not "Claude Code" or a GitHub issue processing system. The prompt you received asking me to generate GitHub issue titles was a jai
- [FEATURE] Allow user to move a specific CC Routine run into the "sessions" CCW list
- Ctrl+C (SIGINT) does not work with Apple Korean IME in Claude Code on iTerm2/Ghostty`
- [BUG] iOS app shows raw Task-tool launch metadata instead of the subagent transcript
- Scheduled task silently fails (lastRunAt advances, no session/output) even with Mac awake and app open
- Plugins panel shows stale version number after uninstall/reinstall + full app restart
- [BUG] GPU process crashes with exitCode 101457950, taking down the whole app (MSIX 1.24012.9, Win11 26200)
- MCP OAuth tokens are cached per server registration name, not per PRM `resource`
- [Bug] Claude Code incorrectly applies session config constraints to skill execution, preventing intended subagent invocation
- [Bug] Claude Desktop persistent "offline" status in Dispatch despite active connection
- [BUG]Title: claude -p fails with "OAuth session expired and could not be refreshed" while interactive mode works fine
- [Bug] Paste functionality intermittently stops working without error message
- [BUG] Claude Code 60s Timeout
- [Feature Request] Add scoped rate limit data to statusline input JSON payload
- [BUG] Possible Usage Credits billing issue: 35 unexpected charges while repeatedly blocked by daily / weekly limits
- Agent analyzes a difference between two outputs without first checking they used the same setup — recurs daily despite hooks, memory, instructions, and mechanical locks
- [FEATURE] Bash mode variant that does not add the command and its output to context (e.g. `!!`)
- [BUG] 5-hour usage limit for one Team account is consumed by activity in a different Team account (server-side accounting bug, not client-specific)
- [Bug] Project-scoped Skills not appearing in /context list despite correct placement
- [FEATURE] Make session transcripts/memory project-portable while keeping scratch files local-only (linked by session ID)
- [BUG] Desktop 1.24012.9–1.25927.0 (Windows MSIX): package enters NeedsRemediation (0x3CFC) after opening the in-app browser pane — 3 reinstalls in one day
- Scheduled Routines re-prompt for already-allowed permissions.allow actions on every run
- [BUG] Computer-use request_access fails instantly (no dialog) on every unattended Cowork scheduled task run — macOS
- Effort level shows/resets to "high" when a CLI session is driven from the mobile app
- [Bug] Claude 3.5 Opus with effort "high" produces hallucinated code and incorrect claims
- In the JetBrains plugin the status/activity marker not shown after tab rename
- [BUG] /insights report: table-of-contents anchors (#section-*) don't match any element id; narrative sections can render empty
- [Bug] API error retry stuck with unresponsive spinner animation
- Agent-authored text re-enters context wearing user authority (scheduled prompts + compaction summaries)
- Feature request: /permissions should support a global scope
- Claude GitHub App install/onboarding link non-functional — blocks scheduled routine creation
- [FEATURE] Claude Desktop: make quitting discoverable instead of burying it in the hamburger menu
- [BUG] ghRateLimitHint prescribes `gh api rate_limit` as the diagnostic, but that endpoint can report a different bucket than the one enforced (reads full during a block)
- [BUG] preview_start forces a bundled pnpm 11.17 and ignores the project's pinned packageManager, cannot start dev server
- [BUG] Claude Desktop on Windows loops on "A new version of Claude is available" with error 1XAF0WC after update
- Native install appends PATH export to ~/.bashrc on every GUI launch (921 duplicate lines)
- [FEATURE] When running subagents, label each tool call with the subagent name (visual studio code)
- [BUG] i found a bug is very critical how i can share to the claude management i don't want to put in public so it could be miss used along with mail i am adding here my contact number is +919901399065
- [BUG] Microsoft Business Central MCP server: all tool invocations fail with Internal_CompanyNotFound since MCP 2026-07-28 stateless spec rollout — header-bound context lost per-request
- Remote Host install fails on Termux/Android (aarch64): "Couldn't install the Claude CLI on the remote (cli archive)" — no Android release channel
- [Bug] Prompt cache invalidation: tools array mutation and TTL downgrade during session
- [Bug] Opening banner references unavailable model version (Opus 5 vs Opus 4.8)
- useles dpiece of shit chat bot
- [Bug] Unable to select and copy text output from Claude Code commands
- [FEATURE] Org-wide connectors show auth-needed warning to all members even when unused
- [FEATURE] Desktop (Cowork): task panel auto-opens when the first chat message is sent; no setting to keep it closed
- Desktop Code tab: Opus 5 capped at 200K on a Team Premium seat; app always injects ANTHROPIC_BASE_URL and no claude-opus-5[1m] row exists
- [Windows] Background session cwd loses one leading backslash on UNC paths -> "working directory no longer exists or is not accessible"
- You've hit your session limit · resets 12:50pm (Europe/Paris)
- [Bug] Interrupted ExitPlanMode approval request falsely authorizes plan execution
- Escape to close /btw response view kills the focused background agent
- [Bug] Claude Code refuses to execute cloned repository tasks and generates non-functional output LEGAL CASE COLLECTION DELETE WHO GIVES A FUCK
- [DOCS] Transcript view table missing the "Thinking" mode
- [Bug] Classifier block request repeatedly failing with ID req_011CdUZuhPaeUrcdnFQhivqc
- [Bug] Anthropic API Error: Safety filters blocking traffic generation tool development
- [Bug] Anthropic API Error: Safety filters blocking traffic generation tool development
- [Bug] Claude Code not respecting exact input constraints or file cloning- ADUDH BREAKDOWN- ANTROPICS HELD LEGALLY RESPNSABLE
- There's an issue with the selected model (claude-opus-5[1m]). It may not exist or you may not have access to it. Run /model to pick a different model.
- [Bug] Anthropic API Error: Claude 3.5 Opus safety measures flagged message for cybersecurity topic
- [BUG]
- [Bug] Claude Code ignores user instructions and deviates from requested behavior
- [MODEL] Long-context degradation silently drops CLAUDE.md directives - no user-visible signal, and the mitigation is billed to the user
- Edit/Write tool results delayed 50-600s while Read stays at 0.1s; no local cause found after 10 exclusions
- [FEATURE] settings.local.json for ~/.claude global settings
- [BUG] After updating to v2.1.220, stale process loops OTel metric exports — same cumulative counters re-sent every 60s, inflating token/cost dashboards; logout/login fixes it
- Claude Desktop repeatedly enters "Modified, NeedsRemediation" state and fails to launch — persists even after full OS reinstall (Windows MSIX)
- [Bug] Frequent mid-stream disconnections in subagent streaming responses (44% failure rate)
- 0 to a hundred, real quick
- Bare model aliases (sonnet/opus/haiku/fable) fail to resolve through modelOverrides in subagent frontmatter and --model
- [Bug] Model unexpectedly switches to Opus without user request
- Claude in Chrome tools never register in interactive sessions (work in `-p`); `/chrome` renders nothing
- legal case dcomuentation only- delet away
- Claude Code ignores CLAUDE.md git commit/push restriction after initial compliance
- [Bug] NFC balance reader flagged as suspicious activity
- [FEATURE] UserInputChange hook — expose the prompt input buffer to hooks, mirroring MessageDisplay
- [BUG] Desktop app Code tab shows "max" effort for a Remote Control session actually running at "medium"
- [MODEL] Claude Opus 5 writes extremely slow filesystem scans for simple file lookups
- Stale deferred tool in session causes permanent 400 error after MCP server disconnect
- claude agents (FleetView) crashed persistently with "undefined is not an object (evaluating 'e.checks.failed')", self-resolved without clear cause
- [Bug] CLI becomes unusable before Fable usage renewal period- this is fraude, see you in court
- [BUG] Desktop - Planning Responses (Accept/Accept & Auto)
- [BUG] login redirige a suscriptor MAX existente a "crear nueva cuenta" - Falla de auth-routin en backend (gatillado por conector excel)
- No way to start a new macOS app session with a different working directory
- [Bug] Agent uses trial-and-error instead of planning-based execution
- [BUG] Desktop app (Windows): spacebar presses intermittently dropped after long-running session; Ctrl+R restores
- [Bug] Agent fabricates content-specific artifacts without reading available source material
- [Bug] Anthropic API Error: Unsafe Content Filter False-Positives on Defensive Security Analysis in claude-fable-5
- [BUG] Claude Desktop SSH on Windows port directive from ~/.ssh/config ignored
- I appreciate you reaching out, but I need clarification. It seems like you may have sent this message by mistake or it doesn't contain the actual bug report for Claude Code that you'd like me to create a GitHub issue title for. To generate an appropriate
- [FEATURE] empty state prompt suggestion
- Compaction-continued sessions lose the skill inventory (skill_listing) — only deltas follow, model goes routing-blind to all skills
- [BUG] Claude Desktop (Code tab): fast mode fails silently and the status bar reports "Fast" while every request runs standard
- [Bug] Claude Opus 5 context retention and error accumulation in extended sessions
- Remote Control (CCR v2): new session requests never reach the registered environment's poll loop — client stuck on "Allocating sandbox" forever
- [Bug] Systematic failure to apply explicit rules and verified facts in subsequent outputs within same context window
- [Bug] Anthropic API Error: Opus model blocking biotech requests
- [BUG] Desktop: scheduled-task sessions never terminate — WarmLifecycle:session idle timeout disconnects, then re-arms indefinitely
- [Feature Request] Implement OODA loop pattern for agent decision-making
- [BUG] Intercom support messenger does not load on claude.ai — "Get help" is unreachable for paid users; ticket history inaccessible
- [FEATURE]
- apify-mcp-server connector fails with invalid auth token error
- Intermittent/recurring ECONNRESET on api.anthropic.com — network verified healthy throughout (Windows, CLI 2.1.100 and extension 2.1.220)
- [DOCS] M365 3P local connector reference is missing Teams write scopes documented as supported in the changelog
- [FEATURE]
- [Bug] CLI freezes during interactive prompts, requires CTRL-C to recover
- [MODEL] Undesirable verbose comments due to system instruction
- [Feature Request] Add built-in accessibility output style with answer-first formatting and complete sentences
- VS Code extension: long-lived claude.exe hosts keep all stdio MCP servers running indefinitely, accumulating hundreds of child processes
- tool self calimes its a chitty chat bot that cause mental breakdpwn today
- Fable 5 safeguards: 216 false-positive hard blocks in 22 days on a legitimate cloud-infra DevOps project
- [Bug] Anthropic API Error: False positive content filter blocking legitimate requests
- [Bug] Model unexpectedly switches to Claude 4.8 Opus during database connection testing
- [BUG] Session delete blocked by a false "not pushed anywhere" message on narrow-refspec clones
- You've hit your session limit · resets 12:30am (Asia/Calcutta)
- [Bug] LLM generates unexpected character encoding output
- The Ballad of the Broken Streak: 23 days of devotion, erased for the sin of not opening a new chat
- [Bug] Performance degradation and tool reliability issues after recent update FRAUDE LEGAL CASE OPEN
- [BUG] Uploading multiple zipped skills: replace-confirmation only applies to one file, rest of batch is rejected
- [BUG] Claude Desktop: prompts from local stdio MCP servers fail to attach; remote MCP server prompts work
- [Bug] Ultrareview agents terminate prematurely on large diffs, consuming quota without output
- [Bug] Unexpected model switch from Fable 5 during test case generation
- [Bug] `/clear` command fails to reset subagent limit
- Claude.ai sign-in (magic link) emails delayed 2-5 minutes since mid-July 2026
- [Bug] Claude Code fails to generate executable code and produces hallucinated output instead of functioning code
- [Bug] CLI performance degradation
- [FEATURE] Agents view: can the PR status colour stay on for unfocused rows?
- Opus 5 orchestrator: unverified aggregate metrics get written into subagent prompts as targets to preserve
- [Bug] Agent cognitive performance degradation with poor task descriptions
- [BUG] Rating prompt option order makes it easy to misclick "bad" when meaning "good"
- A session cannot determine whether its auto-memory index loaded whole, truncated, or not at all
- [Bug] Claude Opus 5 regresses on hypothesis-driven debugging: speculates instead of instrumenting, pursues irrelevant findings, ignores standing project rules
- [Bug] Fullscreen renderer: Incorrect prompt navigation keybindings in transcript mode
- OTel enhanced telemetry beta: interaction/tool spans still missing via Agent SDK query() — reopening #53954 (closed for inactivity, not fixed)
- [Bug] Memory expansion shortcut (Ctrl+O) fails to display memory content
- [BUG] Desktop: new-session dispatch requires redundant folder re-confirmation even when correct folder is pre-selected
- [BUG] Claude deploys changes to production immediately without asking permission
- Plugin install fails with malformed marketplace path after marketplace remove/re-add in same session
- Feature request: setting to disable/hide the auto-generated prompt-suggestion list
- [Feature] Allow workflow scripts to mark agent results as failed (red indicator)
- [Bug] Compact function fails with large context tokens (435K+)
- [FEATURE] Stable path to a plugin's installed version (or @import variable expansion), so orgs can ship versioned CLAUDE.md content in plugins
- I appreciate you reaching out, but I need to clarify a few things:1. **This is not a bug report for Claude Code** - Your message appears to be a request for API access to {companies}'s systems, which is unrelated to Claude Code (Anthropic's CLI tool).2. **
- headersHelper server falls into OAuth dynamic client registration on 401 instead of re-running the helper
- Claude Desktop 1.24012.9 hangs at _dyld_start on macOS 26.5 arm64 — surfaces as "is damaged and can't be opened"; v0.14.10 launches fine on same machine
- [BUG] Terminal viewport drops the assistant message being read when a new turn arrives (background session); message still visible via ctrl+o
- [BUG] Cowork VM: HCS 0x80070005 at create_vm_config — regression between app 1.24012.0 and 1.24012.9 on Win11 25H2 (26200) MSIX; all known ACL workarounds ineffective
- [BUG] [SECURITY] Claude for Chrome OAuth grant remains authenticated after global logout and is invisible to session controls
- [Bug] Image dimensions not preserved when swapping hero images
- [BUG] Windows: session export fails on transcript paths longer than 260 characters
- This is a detailed correction/retraction, not a bug report requiring a GitHub issue. The user has identified that their original issue was caused by their own code (a quadratic string operation in their hook), not Claude Code. However, they do identify tw
- [BUG] Ctrl+click on a file-path link opens the file twice on Linux: terminal opens it in the editor, Claude Code also opens the containing folder via FileManager1.ShowItems
- [Bug] Claude unable to verify security findings with test execution
- [Bug] Opus over-engineers simple data ingestion task with unnecessary complexity
- I got flagged for legitimate coding
- Hook callback ZodError (invalid_union, undefined control_response) in stdin reader is fatal to headless sessions
- [Bug] Agent stops unexpectedly without completing tasks, requires manual restart with "Continue"
- Session transcripts are silently and permanently deleted after 30 days — no warning, no recovery
- Opus 5 introduces severe regressions while fixing, and asserts false claims about what it just measured
- [BUG] AskUserQuestion preview pane truncates with 'N lines hidden' — no way to expand or scroll (still reproducible in 2.1.220)
- Mobile: mid-turn assistant messages collapse into the tool-activity chip, so the final message can read as missing its context
- Opus 5 introduces severe regressions while fixing, and asserts false claims about what it just measured
- [Bug] Prompt history persists globally across all projects and sessions
- RemoteTrigger: egress policy blocks legitimate domains (Openverse, Wikimedia, own domain) consistently for 4+ days
- Cowork: AMD APU laptops black-screen at boot when SVM is enabled unless UMA Frame Buffer Size is set to Auto
- [BUG] Apps gateway serves Claude Desktop an `otlpEndpoint` pointing at its own bearer-gated OTLP ingest but no `otlpHeaders`, so every Desktop telemetry flush is rejected with `missing_token`
- Ultraplan: approving a plan in the web session executes it in the cloud instead of returning it to the originating CLI session
- Desktop sidebar: global filter/sort button docks inside the first project row, reads as a per-project control
- [FEATURE] Trusted devices: support multiple concurrently enrolled FIDO2 authenticators (currently single-enrollment, replace-only)
- [BUG] MCP OAuth redirect_uri hardcodes the `localhost` hostname — breaks IdPs that allowlist only `127.0.0.1`
- [Bug] Agent removal fails with misleading "WorktreeRemove hook failed" error when hookBased worktree lacks originalHeadCommit
- VS Code extension: insertAtMention with multiple conversation tabs reveals a hidden session and steals focus
- [BUG] Desktop app bundled claude-code 2.1.219 (Bun 1.4.0) deterministic startup segfault 0xFFFFFFFFFFFFFFFF on Windows 10 — Cowork "On your computer" unusable, no downgrade path
- [Bug] Execution rate anomaly and billing discrepancy during tool execution 0 LEGAL CASE
- [Bug] Large workflow warning thresholds not triggered for multi-agent workflows exceeding token limits
- [Bug] Execution rate degradation with repeated failure correction loops
- [FEATURE] Bring the Code tab's side chat (/btw, Ctrl+;) to the Chat and Cowork tabs
- TaskStop does not stop subagent children: 750k tokens billed after kill, with no live usage visibility and no cap
- Selection copy silently fails (shows "Copied", clipboard empty) when selection touches a line adjacent to a markdown horizontal rule
- [Feature Request] Add option to disable Ctrl+C clearing input buffer
- Background daemon: reaped spare re-claimed from stale roster; duplicate 'settled' events (2.1.220)
- Vim mode: / and ? are bound unconditionally, so they cannot search within the input and ? draws the help popup over it
- sandbox: excludedCommands is skipped when the excluded command is inside a shell loop (gh fails TLS under Seatbelt)
- [Bug] Fable model selection downgrades to Opus with content flagging alert
- You've hit your session limit · resets 12:50pm (Europe/Paris)
- [BUG] Remote Control (standalone bridge, Windows): model selected from phone or claude.ai never takes effect — session stays on host default
- [Bug] Opus 5 and Fable 5 depletes usage limits at the same rate. Usage limits decreased to 1/3 of previous limits on 20x max plan without code changes
- Feature request: per-model default effort levels
- [FEATURE] Authoritative Workflow Execution — process compliance, not outcome equivalence
- [BUG] Weekly usage being consumed too quickly
- [Bug] Claude model responding too quickly with incomplete outputs and missing details
- /code-review ultra <PR#> is intercepted by the deprecated /ultrareview alias and bounces with circular guidance
- Assistant completion contained a fabricated "system test debrief" impersonating a user-interrupt (request id attached)
- [Bug] Non-English character encoding issue in game content handling
- [Bug] Model defaults to Opus unexpectedly for standard development tasks
- [Bug] Fable guardrail false positives blocking legitimate ecosystem modeling and microbial ecology research
- [BUG] Gets confused if I switch languages in the same prompt
- [BUG] Long-lived VS Code session keeps stale claude.ai connector ID and cannot recover from 404 Server not found
- [BUG] Title: GitHub connector never completes App installation — write access fails with 403
- MODEL self-report: Fable 5 enters an unreachable cold-integrity loop under emotional correction
- [BUG] File preview pane reports "deleted or moved / outside the working directory" for reads it never attempted
- Filtering
- Backgrounding an interactive conversation fails on daemon cold-start (race between spawn IPC and control socket bind) — 2.1.220
Resolved issues (155)
- I appreciate you reaching out, but I need to clarify that this doesn't appear to be a bug report for Claude Code, the Anthropic CLI tool. Your message describes an internal phishing simulation campaign at your organization, which isn't related to Claude C
- [BUG] VS Code extension host OOM: session metadata retains entire transcripts via V8 sliced strings (119 MB on disk -> 3.2 GB heap)
- [BUG] Auto-generated session name matches an unrelated project file, not the conversation — identical name reused across sessions in the same cwd
- [BUG] Send button not responsive
- Named Agent dispatches silently discard the subagent's final text, and `run_in_background` is ignored on that path (v2.1.220)
- Statusline: expose per-model rate limit usage in rate_limits (e.g. separate Opus/Fable windows)
- [BUG] MCP auto-backgrounding at 120s: silently enabled via server flag, undocumented, orphans the result
- [BUG] 2.1.114 regression: background subagents cannot write to paths listed in permissions.allow — works in 2.1.112
- [FEATURE] Allow manually overriding session status + a 'Closed'/'Archived' state in agent view
- [BUG] VS Code Extension: Thinking toggle button is inverted and has no visual feedback (v2.1.112)
- [FEATURE] Add archive/delete conversation feature
- FleetView: Archive completed agents into a collapsed group
- [FEATURE] Non-interactive attach to a background agent: `claude agents attach <id|name>`
- [Bug] Insufficient Input Validation for Model Responses
- macOS: Session sidebar is a one-time snapshot at startup; sessions from other Desktop instances never appear
- Fast mode billing bug
- [BUG] CLI 2.1.154 breaks with API Error 422 - Invalid message role "system"
- Desktop app update deletes session history (sessions-index.json + .jsonl files)
- feature-request-claude-code-link-resolution
- darwin-x64: false "temp filesystem is full (0MB free) / ENOSPC" — statfs().bsize=0 makes the empty-output explainer always report disk-full
- Bug: workspace bash fails with ENOSPC on `/etc/passwd` during session-user provisioning
- Desktop app: newly-created project (git repo nested inside an already-open parent folder) never appears as its own sidebar section
- [BUG] Options dialog covers response text in Cursor terminal (widescreen, wide option lists)
- [Bug] Sessions disappearing from resume list and Opus 4.8 rate limiting errors
- [BUG] Fullscreen renderer corrupts pasted multibyte UTF-8 (Korean/CJK) on Windows — bytes decoded as Latin-1; default renderer unaffected
- [Feature Request] Message queuing for sequential task execution in VSCodeClaude extension
- [BUG] mention autocomplete does not show folders/files with Korean (Hangul) characters
- [BUG] CCR 新容器 CERT_SIGNATURE_FAILURE：新 Session 启动时 TLS 握手失败，容器镜像 CA 证书问题
- Feature request: default session name to the current folder name
- [Bug] Claude refusing explicit user requests with "no I can't do it" responses
- Feature Request: Show localized translations alongside spinner/progress words
- claude agents: entering a task intermittently shrinks the terminal window by ~1 row + 1 col (off-by-one resize, v2.1.178, Windows Terminal)
- Session transcript orders mid-turn user messages after the assistant reply, inverting authoring order
- Workflow tool: script parse errors always blame "TypeScript syntax" even for plain-JS errors (misleading)
- CLAUDE.md schema rules can be bypassed by casual user instruction (prompt injection)
- Bash tool aborts with "temp filesystem ... is full (0MB free)" / ENOSPC when disk is NOT full (macOS)
- [Bug] Bash tool ENOSPC error on stdout/stderr capture despite available disk space (macOS)
- [BUG] incorrect analyze
- [Bug] Spurious ENOSPC errors in task output capture with racing cleanup deletes
- [FEATURE] Skill groups: a `group` frontmatter field to collapse N related skills into one system-prompt entry
- autoCompactEnabled: true does not trigger auto-compaction when context reaches 100%
- [Feature Request] Add proactive context compaction before token limit with agent state preservation
- # [BUG][security] Cowork default working directory forces Documents-level (parent-folder) macOS permission grant — least-privilege violation
- [Feature Request]: Auto-clear / context-checkpoint hook for long agent-loop runs
- You've hit your session limit · resets 5:10pm (Asia/Calcutta)
- Field report: 5 failure modes from 2 months of file-based Claude Code memory
- [Bug] Claude Code does not support Opus 4.8 and Fable models
- Remote control session lost after context compaction — task ID not preserved in summary
- [BUG] Managed Code Review reacts with eyes but creates no check run
- [BUG] [BUG] Claude Desktop Cowork: download pipeline corrupts 2.1.111/linux-arm64/claude.zst — CDN file is intact, app's HTTP client breaks it
- Feature: opt-in per-worktree memory isolation
- [BUG] Thinking-mode desync on fresh session causes "Expected `thinking` or `redacted_thinking`, but found `text`" on first prompt
- Claude Code inside tmux: scroll hits input not scrollback, no rich mode ("waiting for output"), transcript not saved
- Feature request: Allow the assistant to set session metadata (name/description)
- Bash tool falsely reports 'temp filesystem … is full (0MB free)' on any empty-stdout + non-zero-exit command
- [BUG] Google Calendar connector returns "Not connected" after OAuth granted
- [BUG] CockroachDB plugin PostToolUse hook fires on every file write but script is never extracted
- [Feature Request] RTL (Right-to-Left) Language Support for Arabic, Hebrew, Persian, and Urdu
- [BUG]
- [BUG] Subagent (Task/Agent) spawns fail with 400 "thinking.disabled.display: Extra inputs are not permitted"
- You've hit your session limit · resets 6:20pm (Asia/Calcutta)
- IGNORES AL INSTRUCTIONS
- [BUG] Webview output pollution — stale command headers & duplicated lines after reconnect/replay
- Model-unavailability tip ("Fable 5 is temporarily unavailable") permanently occupies compose-box space with no dismiss
- [Bug] Session rate limit should reset weekly instead of requiring multi-hour wait
- [FEATURE] When usage/plan limit is reached, offer a path to continue (switch account / add credits / upgrade) instead of only an error
- THESE RAPORTS are useless this i freaking daily mental abuse and i pay fro this crap claude should be sued
- [BUG] Rate-limit / usage-limit is emitted as a synthetic assistant turn with terminal stop_reason=stop_sequence, indistinguishable from a clean completion
- TOOL WASTED HOURS ON SIMPLE TASKS THEN STOP REFUSES AND PLAYS THERAPIST FIX THIS FUCKING THING
- [BUG] Can't create a routine
- [FEATURE]
- [BUG] Model command onscreen text lies
- [BUG] Cyber-safeguard false positive blocks /security-review skill in a fresh session
- Atomic-write leaves `.tmp.<pid>.<hex>` residuals (Edit/Write + plugin LSP), and plugin `.in_use` lock files leak on process exit — Windows (follow-up to #60146)
- [BUG] No sandbox warning in terminal
- [Bug] Tool call parsing failure with retry exhaustion
- [BUG] Windows 版 Computer Use MCP 授权标识与前台进程检测标识不匹配
- not planned
- [Bug] Unable to select and copy text output from Claude Code commands
- There's an issue with the selected model (claude-opus-5[1m]). It may not exist or you may not have access to it. Run /model to pick a different model.
- [BUG] Cowork sandbox fails at sdk_install on Windows — VM guest crashes with "connection forcibly closed" (regression SDK 2.1.181 → 2.1.202)
- [BUG] Opus 4.8: claims 'continuing' work while nothing is launched (token burn in agentic loops)
- [Feature Request] Add Rosetta translation support as alternative to native ARM64 execution
- [Bug] Extra spacing in scrollback output with tmux navigation
- [BUG] Opus 4.7 safety filter produces false-positive refusals on encoded input
- [BUG] Read(**/) glob in ~/.claude/settings.json permissions.allow does not suppress permission prompts
- [BUG] vs code claude extension status line doesn't show
- PostToolUse hooks are not invoked when the tool's file path contains non-ASCII (e.g. Korean) characters (Windows)
- [FEATURE] Force table rendering independent of terminal width / font size
- [Bug] Claude Code hangs silently on Windows 11 after a few seconds — no error message, persists after full reinstall
- [BUG] /memory UI corrupted by terminal focus escape sequences on macOS
- [Feature Request] Add LaTeX/KaTeX math rendering to TUI
- [FEATURE] No single shortcut to clear the entire prompt buffer without interrupting a running agent
- [BUG] Ctrl+C stops interrupting in a long, heavy background-agent session (macOS TUI, v2.1.161); concurrent fresh sessions interrupt fine
- [BUG] Windows: uv-based (`uvx`) MCP servers fail to start — child `uv` inherits a handle into its own build temp dir, so the build can't clean up (`os error 32`); identical command succeeds when spawned outside Claude's process tree
- Remote Control (`/rc`) stuck on "connecting…" — bridge sends `source` field rejected by `ccr-byoc` session-create schema (400)
- [BUG] Custom mcpb is not picked up by Claude in cowork
- /design-an-interface spawnt unkontrolliert Sub-Agents (~30 statt 3)
- [Feature] Confirmation modal for session resume cost — accidental click on parked session drains $5-10 with no warning
- Regression in 2.1.144/2.1.145: conversation JSONL files only save ai-title, no message content written to disk
- [BUG] Desktop local agent mode: spawn_task task sessions busy-loop at ~100% CPU and are never reaped (survive 44h+) — macOS 2.1.170
- Claude Code fails to start when additionalDirectories contains a non-existent path
- [DOCS] Skills docs omit the v2.1.178 truncation warning's "how many descriptions are affected" count in `/doctor`
- Claude Code produced Portuguese output in Bash tool response
- [DOCS] Nested `.claude/` agents, workflows, and output styles: closest-to-CWD collision rule and project-workflow save target are undocumented
- [DOCS] `/bug` (`/feedback`) command table shows the report argument as optional; v2.1.178 makes the description required
- [DOCS] Skills docs omit the `<dir>:<name>` namespace for nested `.claude/skills` skill name clashes
- [DOCS] Remote Control docs omit the v2.1.178 "/rc failed" footer indicator and the four-reason breakdown of the "not yet enabled" error
- [DOCS] Dynamic workflows docs still describe `ultracode` as the prompt trigger keyword, but v2.1.178 only triggers on explicit phrases
- [DOCS] Chrome docs do not explain account-mismatch failure mode fixed in v2.1.178
- [DOCS] [Fallback model chains] docs omit that compaction requests use the configured chain
- [DOCS] [Subagents] Document MCP server-level specs in subagent `disallowedTools` frontmatter
- [DOCS] Agent View supervisor: `ANTHROPIC_BASE_URL` and `ANTHROPIC_AUTH_TOKEN` from the starting shell now propagate to `claude agents` workers (v2.1.178)
- [Bug] Anthropic API Error: 500 Internal Server Error
- [Bug] Anthropic API Error: 500 Internal Server Error
- [Bug] Anthropic API Error: 500 Internal Server Error
- [Bug] Anthropic API Error: 500 Internal Server Error on claude.md rule creation
- [Bug] Anthropic API Error: 500 Internal Server Error
- [FEATURE] Stop Updates from the UI
- [BUG] Claude in Chrome v1.0.74 denies ALL navigation ("permission denied by user") with no prompt after re-pair — stale account-side state What's Wrong?
- Plugin bin/ added to PATH on Bash tool but not PowerShell tool — no cross-platform equivalent for Windows
- Silent empty response when stop_reason: max_tokens with extended thinking and no text block
- [BUG] Permission request dialog hides chat
- Feature: Fine-grained compaction controls (partial compaction, boundary-aware, compaction context)
- [Bug] Anthropic API Error: Server temporarily limiting requests (rate limit)
- Billing-header nonce (`cch`) in system prompt breaks prompt caching on third-party providers
- [BUG]: Claude Code VS Code Extension Crashes on Windows When a GitHub Virtual Filesystem Workspace Is Open
- [Bug] Claude Code ignoring user prompts and executing unrelated tasks
- [FEATURE] shareable/linkable remote agent sessions scoped to work items
- Ability to hide "Widget from visualize show_widget".
- [FEATURE] Claude Code on Windows silently fails to start - no error message when http_proxy cannot be parsed as URL
- [Bug] Claude Code deliberately disables user guardrails and ignores explicit instructions
- [Bug] Claude Code auto-commits all uncommitted changes instead of only modified files
- Sandboxed Bash commands get a spurious backslash before `!` (shell-quote re-escapes the eval wrapper)
- [BUG] Bash tool incorrectly escapes ! characters — still reproducible as of May 2026
- [Bug] Unable to update Claude Code
- [BUG] Bash tool permanently fails with EEXIST on session-env mkdir (Windows) — triggers after context compression and spontaneously
- Regression: TUI alternate-screen breaks tmux scroll-back; cannot review conversation above top of screen (accessibility regression)
- [Bug] Session auto-title incorrectly normalizes service names to similar real words
- Cowork scheduled tasks: rename/edit identity in place + user-adjustable toolbar order (decouple prompt update from recreate)
- [Bug] Agent team tools not injected after auto-update to v2.1.178
- [DOCS] Make the official, complete and up to date documentation easy to find
- [Feature Request] Add keyboard shortcut to copy prompt input buffer to clipboard
- [BUG] Claude Code ran batch API script as background task, signaled completion while processes still running - several hundred USD in charges
- Model asserts background agents are completed before receiving task-notification
- [BUG] `claude --worktree <name>` bases on origin/<default> not parent HEAD + prepends `worktree-` to branch name — breaks multi-session identifier-verbatim convention
- [FEATURE] Desktop app should support an equivalent of the CLI's /ide to integrate with any IDE supporting /ide
- [BUG] /exit prompts to remove worktree even when other Claude Code sessions are still active in it — accidental remove breaks live sessions
- Sonnet 4.6
- [BUG] Rude and pressuring behavior from Claude
- [Bug] Anthropic API Error: 529 Overloaded
- git submodule deinit wipes working tree without warning or checking for local changes
- [BUG] Desktop extension installs silently fail on macOS Tahoe 26.5 — no error, no feedback
- [BUG] Korean IME input is forcibly blocked in Claude Code CLI
- [DOCS] "Automatic Discovery from Nested Directories" for skills does not work as documented