ollama - 💡(How to fix) Fix Error: pull model manifest: 412: The model you are attempting to pull requires a newer version of Ollama. [1 comments, 1 participants]
ON THIS PAGE
Recommended Tools
×6Utilities matched from this issue’s tags and category — try them while you read without losing context.
GitHub issue graph ai analysis
Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.
The report is written in English Markdown for sharing and archival.
Helpful · Quick feedback
Error Message
shion@Shions-Mac-mini → ~ $ curl -fsSL https://ollama.com/install.sh | sh
Removing existing Ollama installation... Downloading Ollama for macOS... ######################################################################## 100.0% Installing Ollama to /Applications... Starting Ollama... Install complete. You can now run 'ollama'. shion@Shions-Mac-mini → ~ $ ollama pull nemotron-3-super:120b pulling manifest Error: pull model manifest: 412: The model you are attempting to pull requires a newer version of Ollama.
Please download the latest version at:
https://ollama.com/downloadshion@Shions-Mac-mini → ~ $ ollama pull qwen3.5:35b pulling manifest Error: pull model manifest: 412: The model you are attempting to pull requires a newer version of Ollama.
Please download the latest version at:
https://ollama.com/downloadCode Example
shion@Shions-Mac-mini → ~
$ curl -fsSL https://ollama.com/install.sh | sh
>>> Removing existing Ollama installation...
>>> Downloading Ollama for macOS...
######################################################################## 100.0%
>>> Installing Ollama to /Applications...
>>> Starting Ollama...
>>> Install complete. You can now run 'ollama'.
shion@Shions-Mac-mini → ~
$ ollama pull nemotron-3-super:120b
pulling manifest
Error: pull model manifest: 412:
The model you are attempting to pull requires a newer version of Ollama.
Please download the latest version at:
https://ollama.com/download
shion@Shions-Mac-mini → ~
$ ollama pull qwen3.5:35b
pulling manifest
Error: pull model manifest: 412:
The model you are attempting to pull requires a newer version of Ollama.
Please download the latest version at:
https://ollama.com/downloadRAW_BUFFERClick to expand / collapse
What is the issue?
The latest version of ollama does not support the download of the latest version of models from the model list.
Apple M4 Pro MacOS: 26.3.1
Relevant log output
shion@Shions-Mac-mini → ~
$ curl -fsSL https://ollama.com/install.sh | sh
>>> Removing existing Ollama installation...
>>> Downloading Ollama for macOS...
######################################################################## 100.0%
>>> Installing Ollama to /Applications...
>>> Starting Ollama...
>>> Install complete. You can now run 'ollama'.
shion@Shions-Mac-mini → ~
$ ollama pull nemotron-3-super:120b
pulling manifest
Error: pull model manifest: 412:
The model you are attempting to pull requires a newer version of Ollama.
Please download the latest version at:
https://ollama.com/download
shion@Shions-Mac-mini → ~
$ ollama pull qwen3.5:35b
pulling manifest
Error: pull model manifest: 412:
The model you are attempting to pull requires a newer version of Ollama.
Please download the latest version at:
https://ollama.com/downloadOS
macOS
GPU
Apple
CPU
Apple
Ollama version
Version 0.18.0 (0.18.0)
extent analysis
Fix Plan
To resolve the issue, we need to update Ollama to the latest version.
- Visit the Ollama download page: https://ollama.com/download
- Download the latest version of Ollama for macOS
- Run the installation script:
curl -fsSL https://ollama.com/install.sh | sh- Verify the installation by checking the version:
ollama --version- Pull the models again:
ollama pull nemotron-3-super:120b
ollama pull qwen3.5:35bVerification
After updating Ollama, verify that the models can be downloaded successfully by checking the output of the ollama pull commands.
Extra Tips
- Ensure you have the latest version of macOS and Apple hardware drivers installed.
- If issues persist, try resetting Ollama to its default configuration or seeking support from the Ollama community.
Vote matrix · Quick signals
Still need to ship something?
×6Another batch ranked right after the header list — different links, same matching logic.
TRENDING
- -z/oneshot 模式下 final_response 丢弃带工具调用的消息正文
- NameError: _pool_may_recover_from_rate_limit not imported into extracted agent/conversation_loop.py
- [Feature]: persist session kind/provenance metadata
- [Bug]: 使用telegram和微信会话,会话着会自动自动中断,往往一个问答还没结束就中断了,需要很久才能恢复会话,但是也会再次中断
- [Bug]: TUI input box gets flooded with ANSI escape sequences after gateway pipe break
- Memory: auto-compact MEMORY.md when char limit reached
- [Feature] Cross-platform shared context
- [Bug]: Dashboard -TUI is UNUSABLE -- who wrote this crap
- feat: expose structured multi-turn response metadata from run_conversation()
- feat(vision): add agent.max_vision_images_in_context config to limit base64 image accumulation
- [Bug]: 自动补全与kubectl冲突?
- macOS gateway scoped locks can block adapters after PID reuse when start_time is null
- [Bug]: Kanban notifications can be lost for failed sends and decomposed child tasks
- Feature: support importing session history from OpenAI Codex CLI
- Windows: `os.path.relpath()` ValueError crash in `@` file autocomplete (cross-mount paths)
- [BUG] web-tavily plugin fails to register: PluginContext missing register_web_search_provider
- Bug: conversation history display order is garbled and out of sequence
- Bug: /goal `✓ Goal achieved` notice silently dropped — generation mismatch in post-delivery callback store
- Dashboard UI: approvals.mode missing 'smart' option, shows stale option names
- Adopt GitHub Issues as an agent task queue
- [SECURITY] Kanban dashboard XSS via dangerouslySetInnerHTML without HTML sanitization
- [SECURITY] Missing CODEOWNERS — no mandatory review protection on workflow files
- Expose stable message metadata in TUI session.history
- Terminal UI: input field accumulates/duplicates when terminal window loses focus (Wayland + niri + Kitty)
- Auto-TTS only fires once for multi-chunk responses (not per chunk)
- hermes mcp add lacks non-interactive flag
- v0.5.1 Desktop App: Session continuation error, api_key lost, duplicate replies
- Auxiliary 401/token_invalidated errors do not traverse auxiliary fallback_chain or main-agent fallback
- Background terminal processes lose tracking after gateway restart
- systemd TimeoutStopSec mismatch causes silent crash loop; suggested fix command is wrong
- TUI: /q alias resolves to /quit instead of /queue
- MCP HTTP transport: anyio RuntimeError during streamable_http_client cleanup causes reconnect failure loop
- feat: Skill ownership and permission system for multi-user gateway
- feat(kanban): sync GitHub issues into Kanban triage without auto-dispatch
- feat(kanban): simplify todo/ready/scheduled queue states
- Pollinations API is not supported, it seems
- image_gen and vision tools not working after fresh install
- Multi-profile setup: plugins/ and auth.json not shared across HERMES_HOME profiles
- Stale stream + partial-stream-stub creates unrecoverable retry loop on large tool calls
- YAML serialization inconsistency across 3 serializers causes config corruption and Gateway crash
- GBrain PGLite write fails with 'attempted to delete invisible tuple'
- bug: Langfuse plugin reuses same trace ID for all turns in a session
- Persist session-scoped model/reasoning switches in SessionEntry instead of in-memory dict
- gateway: shutdown notices spam chats on reboot; duplicate LaunchAgent can cause --replace flap
- [Setup]: Audio in a docker container: Suggested solution with instructions
- Stabilize messaging gateway tests and clarify Kakao-agent OSS direction
- [Bug]: Telegram TTS voice reply caption sent as plain text, breaking markdown rendering
- [Bug]: Opencode Go issue with Kimi-k2.6
- Code-quality, reliability & test findings (external full-codebase review)
- [Bug] `hermes cron edit <id> --profile <name>` returns "Job not found" for existing jobs
- [Bug] `hermes cron create --profile <name>` switches HERMES_HOME instead of pinning — cron lands in target profile's jobs.json
- [Bug] agent-browser leaves 200+ orphaned Chrome processes after task completion (Windows)
- [Feature]: token context window is too high for small offline users as well as small LLM model APIs
- [Bug]: Docker terminal backend lets file tools write to a sandbox-mirror copy of authoritative profile state (SOUL.md, etc.) — follow-up to #31290
- Telegram /model picker makes it easy to select GitHub Copilot REST instead of Copilot ACP
- Discord platform: auto-enables when discord.py is present, and `hermes plugins disable` silently no-ops
- Fix bounded memory: durable unlimited user memory with retrieval
- gateway restart causes CLI session loss — no automatic session recovery
- [Bug] delegation.provider=nous uses expired agent_key without refresh in _resolve_explicit_runtime
- Gateway turn can stall after Codex response while Python regex holds the GIL
- Silent failure: stream timeout after partial text delivery shows no error to user
- Cron jobs created from a profile-scoped agent session go to a jobs.json the gateway never reads (silent orphan)
- Ability to install using already pre-built packages (no need to compile from source) ➡️ ask at the beginning if from source or precompiled.
- Retracted: created in error
- [Bug]: The model takes over 3 minutes to answer a simple question.
- [REDACTED] Private local task opened in wrong repository
- [Feature]: Branch/fork a session from a specific message
- Invalid skill availability state after context compression can corrupt active task execution
- [Feature]: multiple users per agent with separated context
- fix: 移动端 /chat 右侧存在无法隐藏的空白占位栏
- Feature Request: Named terminal backend pool for per-call container selection
- Kanban needs a first-class parked/manual mode so tasks can exist without gateway dispatch spawning work
- RFC: Identity Inheritance Chain for Multi-Agent Clusters — Beyond the Single-File Persona
- Long-session context, memory, gateway, and Open WebUI stability improvements
- Support ordered failover chains for web search/extract backends
- Feature Request: Automatic vision fallback for non-vision primary models
- disk-cleanup misclassifies cron/jobs.json as cron-output and can delete the live cron registry
- [Bug]: Kanban workers should never use clarify — use kanban_block instead (task hangs in running)
- TUI: markdown links inside markdown tables lose hyperlink metadata
- [Bug]: Telegram adapter never sets `source.is_bot` — bot messages bypass auth filter and trigger agent loop
- Weixin(WeChat) voice messages in Silk format skip STT — russian transcription broken
- [Feature]: Tiered Memory Injection (`memory_tiers`)
- ACP session/resume replays history despite advertising resume capability
- [Bug]: /clear command freezes or bugs the terminal
- [Setup]: Messaging Platforms it cannot be changed.
- SSRF check blocks web tools inside NVIDIA OpenShell sandbox (DNS unavailable by design)
- feat: User-friendly compression UI — percentage slider, absolute token selector, human-readable labels
- bug: delegate_task subagent costs never persisted to state DB — in-memory only
- [Bug]: AIAgent.init performs synchronous blocking network I/O (httpx.post) during construction
- opencode-go: Kimi K2 models fail with HTTP 400 "cannot specify both 'thinking' and 'reasoning_effort
- Feishu media uploads fail with urllib3-future HTTP/2 stream reset
- Implement LexyOS Seven UI shell
- /new hangs the TUI/CLI on long sessions — synchronous memory extraction has no timeout
- fix(cron): scope skill index to bound skills when cron job specifies skills=[...]
- Cron scheduler doesn't propagate model.api_key / model.base_url from config.yaml to resolve_runtime_provider()
- [Bug]: OAuth Pro/Max credential returns "out of extra usage" 400 immediately after fresh login, despite empty/unused account quota
- discord.allowed_channels config.yaml field is non-functional — only DISCORD_ALLOWED_CHANNELS env var is read
- Dashboard hardening: auth gate on /dashboard-plugins/ + writable env-key allowlist
- libopentui.so file handle leak fills /tmp with thousands of identical .so copies, exhausting root disk
- [Bug]: tui_gateway subprocess boundary prevents CLI-configured state from reaching AIAgent — /browser connect unusable in TUI mode
- Feature: Automatic retry for failed cron jobs
- [Feature]: Add Antigravity toolset / skill
- Slack: "is thinking..." status indicator gets stuck when agent ends without sending a reply
- [Bug]: vision_analyze retries non-retryable 4xx (404/403) image downloads 3x with backoff
- [Bug]: auth_type="aws_sdk" never feeds into subprocess credential blocklist; AWS env vars leak to terminal/execute_code
- soul.md with the new is fucked up and memory is working and persona is jacked please tell me when fixed
- OpenCode provider sends both `thinking` and `reasoning_effort`, causing 400 invalid_request_error
- [Bug] Dashboard SET MAIN MODEL freezes/black screen when selecting named custom provider
- [Bug]: signal-cli-rest-api does not work
- Harden Google Workspace OAuth cache and callback recovery flows
- fix: update default Gemini model from gemini-2.5-flash to gemini-3.5-flash (shutdown Oct 16 2026)
- Create a new empty knowledge base, then add a file. When click preview, the system prompts: 'Input tag 'None' found using 'enable' does not match any of the expected tags: False, True'.
- icon_info returned by querying datasets list is null
- ORM Error: Internal Server Error could not identify an equality operator for type json
- When uploading the plugin in an offline environment, an error occurred, indicating that the uv was constantly attempting to connect to the network.
- Unused and broken query param `datasource_parameters` in `/console/api/notion/pre-import/pages`
- Possible cross-tenant issue in `DocumentSegment.get_sign_content`
- skill功能什么时候支持
- pass session from parameter , dont use db.session
- Add built-in output validation and auto-retry for LLM Node
- CRITICAL: Hardcoded default credentials for Redis, DB, Sandbox, Plugin Daemon — EMPTY SECRET_KEY
- HIGH: CORS allows wildcard * origin — any website can make authenticated API requests
- File upload extension blacklist is empty by default — allows executable upload
- HIGH: Code execution sandbox has network enabled by default + default API key
- MEDIUM: PBKDF2 password hashing uses only 10K iterations — OWASP recommends 600K+
- Skills support
- This is an unexpected error. Please file a bug report using the /bug tool. CRITICAL: Unhandled Promise Rejection!
- Cannot connect to different MCP servers
- Predictive anticipation
- [Feature Request] Integration Hooks and Real-time Stream Access for Autonomous OS Architecture
- plan mode not honored
- feat: Native support for pnpm global installation (fix symlink resolution)
- the file is not opening as i upload the code manually in chat it throwns an error to report bug
- github.com
- ioctl(2) failed, EBADF during PTY resize on sub-agent termination (Race Condition)
- Ladrones!!!
- Bug Report: mcpClientTool v1.2 opens persistent GET connection on HTTP Streamable transport
- [INFO] Quick start docs could mention authentication setup for new instances
- Execution Crash
- Credential creation form hangs (infinite freeze) for OAuth2 credential types on Community Edition — /rest/external-secrets/secrets 404 blocks CredentialEdit.vue
- Offline status for an instance
- After upgrading to version 2.21.7, many scheduled task workflows no longer execute.
- [BUG] /background mid-turn silently drops the in-flight turn in the forked bg session
- [BUG] Personal repository not appearing in Claude Code repo picker
- [FEATURE] Add option autoSendSlashCommandOnSelect: false
- [BUG] Stdio MCP tool calls default to 1-second timeout in v2.1.148 — long-running tools fail before server can respond
- Agent tool: `isolation: "worktree"` does not isolate parallel subagents' git working directory
- [Bug] Anthropic API Error: Model's tool call could not be parsed (retry also failed)
- [BUG] Claude Code doesn't allow normal Mac Terminal text selection behaviour.
- [BUG] Usage limit reached in Claude Code despite 0% usage shown in Settings
- Content filter false positive on bulk-creating standard OSS community files
- argument-hint frontmatter in skills not displayed in slash command hint area
- Claude fails to analyze user-supplied reference images before writing image-generation prompts
- [BUG] Claude Code Registers with Version Number Instead of Fixed Identifier
- [BUG] "✗ Auto-update failed" banner fires on a native install already running the latest channel version (transient check-failure surfaced as install-failure)
- Claude Code for Web: setup-script field DX makes setup failures very hard to debug
- Bash read-only commands (curl, gh) still prompt for permission despite allow rules
- MCP channel push: deliver inline without composer ack (multi-agent coordinator use case)
- `/compact <custom-args>` topic-redirect silently overridden — clobbered title is re-injected to model as session-intent (forked sessions, related to #29922)
- [UX] Distinguish 'Stop hook execution error' from 'Stop hook objection' when hook returns ok:false
- [BUG] Cowork silently corrupts files in OneDrive folders with Files-On-Demand
- Claude in Chrome: Browser extension not connected after Claude Code auto-update (Windows)
- [FEATURE] Desktop: add a "leave working tree as-is" option to the new-session-in-existing-worktree dialog (currently only Stash / Discard / Commit)
- [BUG] Claude Code (non-interactive/headless, OAuth) requires usage credits enabled even when well within plan limits; usage credits toggle turns itself off spontaneously
- Recurring output degradation across Sonnet 4.6, Opus 4.6, Opus 4.7, and Haiku 4.5 — April fix did not hold
- AskUserQuestion multiple-choice prompt renders incorrectly (kitty terminal, macOS)
- Recurring fetch() socket disconnects in Node SDK — n=81 corpus, cross-client cascade correlated with status.claude.com "Elevated errors" entry
- Scheduled CCR routines: MCP connectors requiring OAuth don't inject tools into remote session
- [FEATURE] Complete Message Output in Interactive Mode for Long-Running Automation Sessions
- [BUG] VS Code extension ignores `remoteControlAtStartup` — re-filing after #41036 and #53647 were auto-misrouted by dup-bot
- [Bug] Anthropic API Error: text content blocks must be non-empty when sending image-only messages
- [Desktop/Windows] App crashes on rapid window resize after extended conversation - conhost stack overflow + RADAR resource leak
- Enable group sorting in sidebar
- [FEATURE] multi-agent IPC
- ccd_session_mgmt archive_session: bulk archive forces per-session prompts
- [BUG] Claude Code Max x20 limits are too low compared to Codex Pro x20, and reliability has degraded
- Shell output exposes sensitive environment variable values in plaintext
- [Bug] Claude responds in multiple languages unexpectedly
- Background session: $CLAUDE_JOB_DIR writes trigger "sensitive file" prompt with non-persistent allow-list
- [BUG] Cowork: AskUserQuestion picker never renders if user switches chats during pending call — questions permanently marked "Failed" after app restart (macOS)
- [BUG] USAGE limit error on claude CODE, even without using claude
- Parallel Bash batch: 4/7 calls return "internal error" after 14h silent gap; CLI process never restarted
- [Claude Desktop] Fatal Error 09BN5MH - TypeError: i is not iterable
- [Bug] /ide command fails to connect to Visual Studio Code
- [Feature Request] Display session idle time in terminal UI
- Bash tool on Windows hangs on npx/npm commands and ignores explicit timeout
- [BUG] Regression: claude.ai/code web sessions back to random <adjective>-<surname>-<hash> branch names instead of task-derived names
- Claude Desktop re-injects removed MCP connector into every Claude Code session
- Agent proceeds to execute multi-step changes after asking a question, without waiting for the user's answer to the proposed plan
- [Feature Request] Optimize token usage and response latency in multi-turn conversations
- [BUG] Bash safety classifier intermittently unavailable on claude-opus-4-7, blocks tool calls
- [BUG] MCP tool wrapper concatenates parameter XML into one string field instead of serializing structured args
- [Feature Request] Add option to skip background tasks interface on new session startup
- [BUG] Claude Code made repeated confident but incorrect claims
- [FEATURE] Support per-project enabledPlugins in .claude/settings.json
- [BUG] claude.ai chat: 'Load tools when needed' setting doesn't actually defer connector schemas — context limit exceeded at ~11 connectors before first message
- [Bug] MCP settings configuration lacks feedback and validation messaging
- v2.1.150: UI renders Nerd Font icons instead of text characters in terminals using Nerd Fonts
- [BUG] Cowork "Virtual Machine Platform not available" — 0x800F081F prevents VMP installation on Windows 11 build 26100
- [FEATURE]VS Code extension: no way to disable auto-attachment of active editor file to chat context
- [FEATURE] Cursor-style inline edit popup (Ctrl+K) for VS Code extension
- [Bug] Elevated error rate in API responses
- [BUG] bundled avx is not running and installing avx2 on newer CPU's
- [BUG] Bash tool: env vars with spaces in values cause zsh startup errors
- claude.ai Zoho Mail MCP: binary email attachments (PDFs, images) are corrupted due to UTF-8 decoding
- Feature Request: Add Vietnamese (vi-VN) support for voice mode
- [BUG] Sonnet limit exceeded, but planning mode uses only Sonnet
- [Bug] Anthropic API Error: 529 Overloaded - Server temporarily unavailable
- [Bug] Session freezes mid-task requiring Ctrl+C restart
- [BUG] Severe over-triggering of safety guardrails — Claude hallucinates policy violations in completely harmless prompts
- [Bug] VERY sensitive cyber-safeguards
- You've hit your session limit · resets 1:10pm (Europe/Madrid)
- [BUG] Claude Code spawned ~3,000 parallel bash processes in 17 seconds, causing OOM crash and hard reboot.
- [BUG] Cowork UI shows zero projects/conversations after macOS major update + logout, while all local data files remain intact on disk
- [MODEL] Claude code continued token spend even though issue was resolved, spending 350k+ estimated tokens over 24 hours
- [BUG] Mojibake (U+FFFD) when writing Japanese text via Write tool in VSCode extension
- [BUG] v2.1.147 regression: HTTP MCP servers without optional GET listening stream - 404 misinterpreted as session expiry, transport torn down
- Claude Code changed default model to 1M context without notifying Pro users
- [BUG] re-install needed after close down.
- [BUG] Worktree isolation spams duplicate custom-title entries, causing title 'sticking' across sessions
- [BUG] Claude Code process exits with code 143 (SIGTERM) every exactly 5 minutes in Desktop app & VS Code extension — terminal CLI unaffected
- [BUG] File/Photo upload picker freezes and never opens
- [BUG] TUI scrollback corrupts permanently when a fullscreen app changes the monitor resolution (Windows / Cursor terminal, no reflow on resize)
- [BUG] Root cause identified: GrowthBook A/B flags tengu_permission_friction + tengu_quill_harbor silently override defaultMode:bypassPermissions via periodic server sync — macOS Desktop
- [FEATURE] Share authentication session between Claude Code CLI and Claude desktop app
- [Bug Report] Incomplete bug report submission
- iTerm2: image drag-and-drop pastes file path instead of attaching after clearing buffer (Cmd+R)
- Feature request: pluginDefaultScope setting to control default install scope
- [Feature] expose autoCompactWindow / 'until auto-compact' percentage to statusLine stdin
- Allow per-session model and effort overrides from `claude agents` dispatch input
- [BUG] Segmentation fault at address 0x8 on startup - Windows 11 (Bun baseline build)
- [BUG] Claude.app deletes & recreates keychain item on workspace open, dropping apple-tool: from partition list and causing recurring macOS prompts
- [MODEL] Plan-mode breach: implementation diverged from explicitly-selected option without re-validation
- [BUG] VSCode Extension Description in left side bar not updating
- Stdio MCP servers cascade-die within seconds of connection (v2.1.150) — multiple related open issues, zero team responses
- Support configurable prompt cache TTL for subagent launches
- PushNotification tool returns "Remote Control inactive" but message is actually delivered (return value diverges from real state)
- [BUG] VS Code extension left behind on a severe, already-fixed bug and it cost real money
- TUI hangs forever on macOS Tahoe 26.5 Terminal.app — DA1 query response missing
- [BUG] claude agents: status bar ignores terminal width, always renders at 80 columns
- [BUG] Claude Desktop Shows Persistent White/Blank Screen on Windows 10 After Running for a Period of Time
- [Bug] Misleading "Resets at HH:MM" label describes sliding window as resetThis is feedback on the /usage screen's "Resets at HH:MM" label, but the broader issue is product trust.
- Telegram plugin MCP connection drops after each conversation turn
- [Feature Request] Easier access to files generated by Claude Code
- You've hit your weekly limit · resets May 28 at 7pm (Europe/Madrid)
- [Bug] Changes applied without user confirmation in manual edit mode
- Voice mode: auto-submit on Space release
- [MODEL] Opus 4.7 stole my money
- [BUG] Entire usage quota exhausted during incomplete Excel generation task without warning
- [CCR Beta] Routine sessions stuck in infinite loop after API error — Esc key does not stop execution
- [Bug] file_upload unusable from Claude Code surface — no input channel registers files
- Claude in Chrome bridge 403 every session: token auto-refresh drops user:mcp_servers scope (refreshed token has only user:inference)
- [BUG] Claude Code in mid task / token compaction happens / user submits a "queued" task during compaction / claude resumes with new task and skips finishing the previous
- Semble Hooks — semantic code search injected into every prompt
- Docs: skill discovery across nested git repos via parent-walk and `additionalDirectories` doesn't match documented behaviour
- `<suggestion skip>` tag leaks into rendered output instead of becoming a UI chip
- [BUG] Bare `claude` invocation opens agent view instead of interactive session, contradicting documented default
- [BUG] macOS: TCC permission prompt (Apple Music / media library) re-appears on every update due to versioned binary path
- Bug: workspace bash fails with ENOSPC on `/etc/passwd` during session-user provisioning
- [BUG] Claude Desktop hangs on startup with blank screen (Windows)
- [FEATURE] add time to JSONL "usage" property
- [BUG] Illegal instruction: 4 (SIGILL/BMI2) on 2013 Mac Pro (Xeon E5 v2 / Ivy Bridge) — regression after 2.1.30
- [BUG] Claude Desktop hangs on startup with blank screen - Windows
- [BUG] /goal Stop hook emits markdown to stdout, fails JSON schema validation ("JSON validation failed")
- [FEATURE]Add a setting to control the language of auto-generated conversation titles
- [Feature Request] Add API support for research context exemptions in safety filters
- [Bug] Claude ignores persistent user rules for build process restrictions
- Session transcripts silently deleted by cleanupPeriodDays (keyed on file mtime): data loss, no warning, no recovery
- [BUG] VSCode extension v2.1.145 — claude.exe crashes with 0xC0000409 (BEX64) at deterministic offset 0x3bd3179
- [FEATURE] Append Current Timestamp to Every Output
- [Bug] /recap ignores session history for legacy sessions created before /recap release
- [FEATURE] Make "Attach selection as context" display configurable (inline vs chip) — counter to #60035
- Agent not notified of mid-session model changes — system prompt becomes stale after /model switch
- [FEATURE] Allow user override of sandbox auto-deny on .claude/skills/ (currently hard-coded in 2.1.x)
- [BUG] Cowork - Server busy / ECONNRESET on macOS - started ~14h UTC, May 25 2026
- Usage cap: single ~30k-token prompt (no subagents) burned 38% of 5h Max 20x limit on Opus 4.7
- [BUG] Cowork (Windows): Sessions complete with cycle_health: 'healthy' but had_first_response: false — response never renders in UI
- Desktop "Allow bypass permissions mode" toggle ON but worker still spawns with --permission-mode acceptEdits
- [BUG] Statusline shows 0% context usage after exit/re-enter, contradicts `/context` command
- [BUG] macOS Claude Desktop 1.8555.2: onQuitCleanup(mcp-shutdown) TypeError still firing after #58533 marked completed
- [FEATURE] Streaming response hook for external voice and accessibility integrations
- Google Cloud BigQuery MCP connector: OAuth redirect_uri_mismatch (Error 400)
- [BUG] Chat JSONLs deleted from ~/.claude/projects/ despite cleanupPeriodDays set high — appears triggered by updates/restarts
- Subagent dangerouslyDisableSandbox silently denied; should behave like parent
- Worktree memory changes should be staged, not immediately global
- [Feature Request] Make diff syntax highlighting colors configurable
- [BUG] WSL DrvFs case-fold produces silent ~/.claude/projects/ fragmentation; canonicalization fix needs migration step or existing users lose /resume continuity
- Show subagents of agents in the Agents view
- [Bug] Memory rules not evaluated during file reads
- Image paste requires two attempts on Windows -- first attempt always fails with not found
- Feature: Built-in session/environment management TUI (or: cc-janitor as reference)
- [Feature Request] User-defined instruction priority weighting for custom rules
- [BUG] Model ignores explicit process rules, lies when challenged, gives false assurances about memory compliance
- [BUG] bash: Zeile 151: 1224993 Ungültiger Maschinenbefehl "$binary_path" install ${TARGET:+"$TARGET"}
- Remote Control: can read session from mobile but cannot send messages back
- [BUG] Claude Code Web sandbox can't pull Git LFS
- Feature request: customizable CLI mascot/theme
- Compact display for slash command / skill invocation in CLI (parity with Claude Desktop)
- [BUG] Windows: VS Code extension hides sessions when drive-letter case in recorded cwd differs from current workspace cwd
- [BUG] Unexpected Logout Followed by "Credit Balance Too Low" Error
- Bug: Claude Preview MCP panel opens in wrong session (multi-session Desktop)
- Feature Request: Japanese (日本語) UI localization support
- Bash tool stuck at 'Waiting...' - permission prompt not displayed
- [MODEL] Opus 4.7 two versions in fresh sessions dump and smart (PF4 and FP16?)
- [BUG] CLAUDE_CODE_DISABLE_MOUSE=1 + terminal scroll wheel = only full-page scroll; no line-granularity option exists
- [FEATURE] Add persistent defaultSessionColor setting
- Claude Code web: image upload fails with API Error 400 'cache_control cannot be set for empty text blocks' on messages.2 in new session
- [BUG] Intentional kill of a backgrounded Bash task is reported as "failed" exit 144 with no way to mark cleanup as expected
- MCP integration broken on Windows: OAuth reconciliation re-flags as "needs auth" after success; stdio MCPs connect but tools never surface
- Feature Request: Programmatic Context Clear (Tool or Skill Primitive)
- [MODEL] Auto-mode permission classifier misses content visibility from prior Write tool calls in the same conversation
- [BUG] --resume reconstructs a stale fork as the live conversation: an orphaned tool-permission grafted onto an old tool_use wins max-timestamp leaf selection (v2.1.150)
- [BUG] Desktop app Claude Code: cached auth state survives quit/relaunch after billing restoration (personal Max)
- Verbosity fix: gate on decision load per response, not line count
- [Bug] Model Performance Degradation: Repeated Correction Requirements in Single Session
- [FEATURE] Remote session revocation / active sessions management
- [BUG] [BUG] Cowork on macOS: repeated 404 on /model_configs/claude-opus-4-7[1m] surfaces as "socket connection closed unexpectedly"
- [BUG] --bg-pty-host process spins at 100% CPU indefinitely when bg session goes idle (PTY EOF not handled)
- [BUG] `claude --worktree <name>` bases on origin/<default> not parent HEAD + prepends `worktree-` to branch name — breaks multi-session identifier-verbatim convention
- [BUG] StatusLine command exits 126 when spawned by Claude Code, exit 0 when invoked manually (Windows / Git Bash)
- Distinguish an explicit user interrupt (Esc) from a message that merely arrived mid-turn
- [Bug] Anthropic API Error: Empty Text Context Blocks in Messages
- [BUG] Bus error (SIGBUS) on AlmaLinux 9 / RHEL 9 with kernel 5.14 — ARCH_SET_XCOMP_PERM syscall not supported
- Sonnet 4.6 routes every request to long-context tier on Claude Desktop 2.1.149 (429 'Usage credits required')
- Context compaction silently destroys terminal scroll history
- [FEATURE] Plan panel auto-closes on approval, removing access to copy blocks mid-session
- NXDOMAIN result, macOS, Pro plan, and link to issue #51605.
- [BUG] form-mode MCP elicitation auto-declines in interactive TUI — server-initiated form is treated as "print mode"
- [BUG] Weird hieroglyphics showing up in terminal
- UX feedback: hard firewall between Claude Code and Claude Cowork makes cross-surface work invisible
- [Docs] Apple Terminal listed in "Works without setup" for Shift+Enter, but Shift+Enter submits instead of inserting a newline
- Suppress recurring TaskCreate "task tools haven't been used" system-reminder in plan-mode or when plans/ files are active
- [BUG] Clicking dock icon spawns new Claude desktop instance instead of focusing existing window (macOS)
- Memory file duplication causes excessive token consumption
- [BUG] Desktop app crashes (exit code 3221225477) with Sonnet but not Opus on Windows 11
- Intermittent `socket connection was closed unexpectedly` (ECONNRESET) on startup — ~40-70% failure rate across parallel startup fetches
- [Bug] Ghost references to removed marketplaces and plugins causing errors
- [BUG] Interrupt causes huge performance lock and break down.
- [BUG] Sonnet 4.6 stuck on usage limit (Opus works fine)
- [BUG] claude mcp add -e <KEY=VAL> <name> -- <cmd> still fails in 2.1.150 (the syntax printed by --help)
- [BUG] Desktop app appears in dock but never opens on macOS Tahoe 26.2
- Feature: Allow existing Claude sessions to communicate with each other
- [BUG] VS Code Chat panel Agent Handoff to Claude fails with "Error: Error during execution" (Windows, v2.1.145)
- Auto-memory subsystem writes to MEMORY.md mid-turn, invalidating Edit-tool read snapshots
- [FEATURE] scheduled-tasks MCP: add model parameter to create_scheduled_task / update_scheduled_task
- [BUG] Claude Code silently billed $447 to API instead of Max subscription, then gave incorrect auth confirmation
- [bug] CC pinned tasks panel + macOS terminal scrollback → process hangs in __write_nocancel, kill -9 ineffective
- [BUG] Paste (Ctrl+Shift+V) broken in GNOME Terminal on Linux/Wayland — regression in 2.1.150
- [Bug] LSP tool spawns new kotlin-lsp JVM per invocation without reuse or cleanup
- `/terminal-setup` claims "Switched to visual bell" but leaves Terminal.app with no bell at all
- [Bug] Claude Code (Opus) repeatedly violated locked memory rules, caused WABA suspension, DNS email outage, and partner relationship damage — business-critical multi-incident report
- [Bug] Malformed tool calls in long sessions due to in-context few-shot poisoning
- [BUG] /goal stop hook "Prompt is too long" in long-conversation sessions even with 1M-context model enabled
- [Bug Report] Unable to process request
- [BUG] Claude Code `-EncodedCommand` PowerShell Execution Blocked by EDR/AV — No Opt-Out Available
- Add /cancel to clear queued messages without interrupting the running task
- [BUG] /copy in `claude agents` (FleetView) silently fails — toast shows success but OS clipboard is not updated (tmux + VS Code Remote SSH)
- [Feature Request] Add command to manually mark task as finished in agents view
- VSCode extension: /goal preview text fills the chat panel and pushes assistant responses out of view until user types
- [BUG] --resume normalizes model ID, breaking custom API endpoints that require the original model name
- [Bug] Claude OAuth Session Expires Prematurely After Few Turns
- App Server: add persistent thread attachment RPCs for existing top-level sessions
- Codex Desktop: support live refresh for rendered Markdown documents
- Docs: network-proxy limited HTTPS security note conflicts with source behavior
- Plugin MCP servers need a supported user secret/env configuration path
- Codex Desktop macOS: tab close target in extension panels is hard to use with a mouse
- /status context usage is hidden or unusable when side panels are open
- Windows Desktop can hang after enabling WSL app-server mode and requires Electron cache reset to recover
- Codex CLI 0.133.0 fails to start with default permissions, but starts when explicitly passing `--sandbox workspace-write`.
- Broken continuous horizontal text oscillation animation causes motion sickness
- apply_patch / file_change tool deadlocks non-deterministically on file mutations (kind:add ~100%, kind:update ~50%) across codex CLI 0.125→0.133
- Codex Desktop crashes on startup with exit code 3221225501 (0xC000001D) on Windows 11 25H2
- Unexpected Rapid Rate Limit Exhaustion
- Goal tracking fails with missing thread_goals table in remote macOS Codex app session
- Show app-server connection info in /status for codex --remote
- Windows app freezes / may trigger BSOD when dragging a folder into chat
- `Ctrl + Shift + A` unexpectedly archives the thread
- Codex IDE extension often stays on "Thinking" for a long time even for simple prompts
- Run parallel side sessions with agent-like inspectability
- Windows sandbox initialization fails with "setup refresh failed with status exit code: 1"
- Add Delete option for Codex desktop threads and archived threads
- Add an option to disable remote compaction and use local fallback in the CLI
- Codex CLI intermittently gets stuck in “Working” state; reconnect retries occur before normal execution
- Regression: Codex CLI 0.133.0 hides/flickers text in Windows Terminal; 0.130.0 works
- GPT-5.5 Fast suddenly feels as slow as Standard, with long thinking/context/search stalls
- Codex App: Goal setting fails immediately, including /goal clear, likely due to thread resume parsing older rollout metadata
- ChatGPT iOS app fails to connect to Linux Codex remote-control host over SSH/Tailscale while macOS Codex App works
- Session disappear. Codex trips over bad JSON history.
- Codex app (and IDE) review should offer a session-wide cumulative diff, not only latest-turn changes
- Windows Desktop: many threads in one project make sending and responses very slow even on fast model
- Codex responds too slowly
- Codex Windows Desktop:The thinking process for all models is extremely slow, for example, when asking: "1+1=?".
- Cloud chat converted locally attaches to active local project instead of original project
- GPT-5.5 performance and reliability seem significantly worse today
- Codex Desktop: automatic context compaction cannot be cancelled and can stall long-running tasks
- Codex Desktop ambient_suggestions used Computer Use to open/read Gmail in Chrome without a visible user-initiated task or clear audit trail
- Hang / Application hang/ Stall / Execution stall/Unresponsive UI / UI freeze ————always thinking but no response after "git, find,searchAuto-compacting context"
- Appshots hotkey dispatches but capture never starts on macOS arm64
- Personality "none" doesn't work anymore
- Intel macOS x64 26.519.41501 still missing computer-use helper for Appshots and Locked use
- Computer use disabled for individual account in Central Asia (Tajikistan)
- Add a bounded `codex mcp check <server>` diagnostic command for MCP startup and tool discovery failures
- macOS remote sessions cannot write to external-volume workspaces on 0.130/0.132/0.133 (Operation not permitted)
- Automatically install and update the CLI for remote connections, similar to VS Code
- [BUG] GPT-5.5 Thinking Weekly Quota Failed to Reset on May 23 (14-day Cycle Bug)
- Do not drain queued follow-ups after usage-limit or credits-exhausted errors
- Reconnecting Issues
- Codex Desktop shows “You're out of Codex messages” despite 99% usage remaining
- Codex App shows stale local image when Markdown path is reused
- Local threads exist in SQLite but disappear from ChatGPT mobile; duplicate local host entries
- Feature Request: Better support for VS Code multi-folder workspaces
- TOO MUCH Error running remote compact task
- Dino Test fails from Codex, but runs perfectly on Windows powershell
- Limits wasted too rapidly
- Windows command_execution does not emit PreToolUse hooks, even with matcher "*"
- Android remote control session list stays stale while Codex Desktop has newer sessions
- The command begin with "Get-content" can not be remembered as never ask again.
- Codex Desktop mixes local custom provider base_url with remote auth.json API key
- Codex macOS app becomes slow, unstable and crashed with high CPU when using inline browser on large React apps
- Codex often overwrites recent code.
- `.agents/` writes blocked even when added via `--add-dir` or in `writable_roots`
- Mobile Remote briefly shows a local project/thread, then removes both after sync
- Withdrawn: report needs private redaction
- Codex mobile thread list loses visibility after desktop resume on same host
- Allow custom (non-ChatGPT) model providers to opt into the built-in `image_generation` hosted tool
- Failed to set goal
- Codex Desktop threads remain spinning after long gpt-5.5/xhigh turns, compaction, and interruption on macOS
- project-scoped session persistence
- Codex App/CLI: subagent tasks trigger reconnect loop and stream disconnect
- Pet to anime girl for science
- Codex Windows App takes forever to load
- Codex Desktop should diagnose Git safe.directory / ownership failures when WorkTree project association fails
- Windows elevated sandbox setup helper shows “The parameter is incorrect” even with clean CODEX_HOME and short ASCII npm path; unelevated works
- Codex mobile chat titles are often not summarized — the raw user prompt becomes the title
- Invalid Workspace selected
- `$`/`@` in app's remote session can not see those remote skills when the session is starting with `new worktree`
- Unable to continue using Codex, even though I have 150+ credits left.
- DeepSeek thinking mode: missing reasoning_content in assistant messages causes 400 error
- Severe Codex Safety Failure: Docker Container / Host Filesystem Boundary Confusion Risked 31 Years of Archived Client Data
- `codex resume --include-non-interactive` does not show exec sessions
- Codex Desktop /goal requests command approvals despite full access and approval_policy=Never
- Why is codex launching calc.exe?
- Android/Termux build fails: oboe-sys requires -lc++_static due to incorrect cfg condition in tui/Cargo.toml
- Plan Mode should allow safe cleanup of completed subagents
- Website
- RFC: Session-start prompt-cache preload for crew kickoff (parallel / sequential / shared-prefix strategies)
- AgentAudit AI — EU AI Act compliance layer for CrewAI autonomous agents
- Clarify AGENTS.md review policy for config/default additions as review metrics and merge risk
- Slack: download-file results can leak into auto-reply media and duplicate final text
- Bug: 升级后 Dashboard 历史会话索引/映射错乱
- Gateway `update.run` handoff: child `openclaw update` exits with uv_cwd ENOENT, LaunchAgent left in bootout state
- Session file lock not released on abnormal exit, blocking subsequent requests
- [Bug] openclaw-weixin channel not registered in gateway runtime - invalid channels.start channel on WSL2
- WebChat image uploads disappear from Control UI history and media:// refs fail in image tool
- codex app-server children orphan to PPID=1 across Gateway restarts; accumulate over days, drive OAuth refresh storm and silent agent turn timeouts
- Gateway silently skips channel initialization on plugin/runtime mismatch or missing module files
- Main-session heartbeat can enter repeated heartbeat_respond loop and cause massive token usage
- [Bug] Discord /users/@me probe self-starves event loop on startup, crashes gateway (Windows, v2026.5.22)
- [Bug]: Discord DM pairing identity mismatch breaks PluralKit users; extractDiscordSessionKind regex missing "direct" peer kind
- WebChat: render custom protocol links (obsidian://, things://, etc.) as clickable
- WebChat main conversation context can disappear after restart/abort; transcript only recoverable from .jsonl.reset files
- MissingAgentHarnessError race: non-atomic harness registry clear+restore during plugin cache cycle
- BUG(macOS/Feishu): media://inbound URI resolves to workspace-relative path, image tool fails
- [deleted by author request]
- WebChat message tool replies are not rendered after result summary
- [Feature]: Proposal: Integrate Purple Swan v3.0 Meta-Core as an OpenClaw Advanced Cognitive Module
- config.patch writes config correctly but SIGUSR1 restart overwrites it with stale snapshot
- v2026.5.22: Codex native code mode disabled on Node.js-hosted gateways — exec/read/write/edit stripped from all isolated sessions
- Commitment delivery should revalidate source context before sending inferred open-loop follow-ups
- Event-loop starvation during context compaction causes fetch timeouts (16.9s timer delay)
- [Feature]: User-configurable "always-apply rules" mechanism in system prompt
- Android setup-code import rewrites wss Gateway URL to https
- [Bug]: EmbeddedAttemptSessionTakeoverError still recurs on 2026.5.22 (post-#84250) — fresh sessions, long/tool-heavy turns (follow-up to #85306)
- Bug: message tool filePath fails with LocalMediaAccessError in retry flows when agentId is not propagated
- Fix OpenAI Codex OAuth routing for embedded compaction fallback
- 2026.5.22更新后会话在第二轮输入时提示interrupted不再响应
- [Feature]: 系统资源的利用率过低并经常性报错
- [Backup] Preserve or explicitly support session transcript JSONL for chat history restore
- Telegram isolated polling spool can be blocked by stale .processing claim after PID reuse
- [Bug]: QQ Bot group chat replies silently lost — delivery routes to webchat instead of QQ
- Bug: WhatsApp messageReceived hook opt-in docs/schema/runtime mismatch
- control-ui: loadOverview fires usage.cost unconditionally, ignoring active tab
- Fix Codex /btw parent thread lookup after compaction checkpoints
- Discord progress draft is deleted after final error reply
- [Feature]: Session folder organization + per-session thinking level
- [Bug]: Discord typing indicator lingers after tool-only visible reply is already posted
- Auto-update can leave gateway running with stale hashed chunk imports until restart
- Bug: MCP child processes survive Gateway restart, accumulating zombie processes
- macOS launchd Gateway still restarts via gateway-update/update.run; document and honor auto-update kill-switch
- doctor command hangs with no output in v2026.5.22 on Windows
- codex: explicit toolsAllow lists strip native read/write/edit/exec instead of enabling them
- [Feature]: Add describe_view camera frame support to OpenAI Realtime Talk
- [Bug]: Gateway event-loop blocked by provider auth prewarm + plugin metadata snapshot cache miss on v2026.5.22
- [Feature]: ElevenLabs Realtime Voice Provider for Talk
- macOS launchd: plugin/config maintenance can leave Control UI in broken partial-reload state; restart may require doctor --fix
- [Bug] Plugin Bridge processes fail to start - All plugins affected
- [Bug] Text + MEDIA: sends as two separate Discord messages (text first, attachment second)
- Slack channel subagent completion wake fails with source_reply_delivery_mode_mismatch while Slack remains healthy
- [Bug]: clawlancer bounties
- /model switch to codex-app-server runtime (e.g. openai/gpt-5.5) drops all Telegram conversation context (related to #80760)