repo stringclasses 1
value | github_id int64 3.77B 4.38B | github_node_id stringlengths 18 24 | number int64 75 77.8k | html_url stringlengths 46 49 | api_url stringlengths 56 59 | title stringlengths 1 746 | body stringlengths 1 55.7k ⌀ | state stringclasses 2
values | state_reason stringclasses 4
values | locked bool 2
classes | comments_count int64 0 128 | labels listlengths 0 5 | assignees listlengths 0 2 | created_at stringdate 2026-01-01 16:55:34 2026-05-05 11:59:03 | updated_at stringdate 2026-03-24 00:13:29 2026-05-05 11:59:15 | closed_at stringlengths 20 20 ⌀ | author_association stringclasses 3
values | milestone_title stringclasses 0
values | snapshot_id stringclasses 76
values | extracted_at stringdate 2026-03-24 23:36:49 2026-05-05 12:00:33 | author_login stringlengths 3 38 | author_id int64 346 282M | author_node_id stringlengths 12 20 | author_type stringclasses 2
values | author_site_admin bool 2
classes |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
openclaw/openclaw | 4,069,311,058 | I_kwDOQb6kR87yjMJS | 44,778 | https://github.com/openclaw/openclaw/issues/44778 | https://api.github.com/repos/openclaw/openclaw/issues/44778 | [Bug] Feishu image causes gateway crash - image path not passed to image tool | ### Bug type
Regression (worked before, now fails)
### Summary
Feishu image causes gateway crash - image tool receives wrong path
### Steps to reproduce
1. Configure Feishu channel in OpenClaw
2. Send an image via Feishu DM to the bot
3. Image downloads successfully to `media/inbound/`
4. Image tool is called but ... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-03-13T06:54:41Z | 2026-04-27T04:47:17Z | 2026-04-24T23:47:07Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | DavidShaoYu | 195,272,300 | U_kgDOC6OebA | User | false |
openclaw/openclaw | 4,007,750,406 | I_kwDOQb6kR87u4WsG | 30,827 | https://github.com/openclaw/openclaw/issues/30827 | https://api.github.com/repos/openclaw/openclaw/issues/30827 | [Feature]: Priority interrupt channel for steering running agents from messaging surfaces | ### Summary
## Problem
When an agent is mid-execution (running a tool chain), incoming messages from
Telegram are queued and only processed after the current run completes.
Impossible to steer or interrupt a runaway agent in real time.
## Current Behavior
1. User spawns a complex task via Telegram
2. Agent begins... | closed | completed | true | 4 | [
"enhancement"
] | [] | 2026-03-01T17:09:56Z | 2026-04-27T05:03:42Z | 2026-04-24T19:10:10Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | ekkerdthomas | 62,130,061 | MDQ6VXNlcjYyMTMwMDYx | User | false |
openclaw/openclaw | 4,211,719,908 | I_kwDOQb6kR877Cb7k | 61,880 | https://github.com/openclaw/openclaw/issues/61880 | https://api.github.com/repos/openclaw/openclaw/issues/61880 | [Bug]: video-generate tool hits 5MB limit of media files, which is too small for generated videos | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
In version 2026.4.5, the video-generate tool failed to deliver the generated video because the size of the generated video is over 5MB.
### Steps to reproduce
1. In version 2026.4.5, configure the `videoGene... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-04-06T12:49:25Z | 2026-04-28T05:27:58Z | 2026-04-25T09:46:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chunliu | 2,128,551 | MDQ6VXNlcjIxMjg1NTE= | User | false |
openclaw/openclaw | 3,949,723,950 | I_kwDOQb6kR87rbAEu | 18,598 | https://github.com/openclaw/openclaw/issues/18598 | https://api.github.com/repos/openclaw/openclaw/issues/18598 | [Bug]: macOS Sequoia: OpenClaw isolated Chrome profile cannot download CSV (chrome://downloads entry is unclickable + wrong filename) | ### Summary
Downloads are unusable in OpenClaw’s isolated Chrome profile on macOS Sequoia: a generated CSV appears in chrome://downloads, but the download entry is unclickable (filename + folder icon do nothing), and the displayed filename differs from normal browsers (Neon job id, no .csv extension).
### Steps to re... | closed | completed | true | 15 | [
"bug"
] | [] | 2026-02-16T21:57:04Z | 2026-04-28T04:44:05Z | 2026-04-26T01:18:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | CraigHorne-Hive | 204,533,085 | U_kgDODDDtXQ | User | false |
openclaw/openclaw | 4,115,667,337 | I_kwDOQb6kR871UBmJ | 52,201 | https://github.com/openclaw/openclaw/issues/52201 | https://api.github.com/repos/openclaw/openclaw/issues/52201 | feat: binding-level workspace override for per-conversation context isolation | ## Feature Request
**Problem:** When one agent is bound to multiple conversations (rooms/groups/DMs), they all share the same workspace. Memory and context from one conversation bleeds into another. The only workaround is duplicating the agent with a different config, which fragments identity and creates config sprawl... | open | null | false | 2 | [] | [] | 2026-03-22T10:53:39Z | 2026-05-04T20:23:21Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | efe-arv | 259,833,796 | U_kgDOD3y_xA | User | false |
openclaw/openclaw | 4,014,522,524 | I_kwDOQb6kR87vSMCc | 32,496 | https://github.com/openclaw/openclaw/issues/32496 | https://api.github.com/repos/openclaw/openclaw/issues/32496 | [Feature]: Support frequency_penalty and presence_penalty Parameters for OpenAI-Compatible Providers | # Feature Request: Support frequency_penalty and presence_penalty Parameters for OpenAI-Compatible Providers
## Summary
OpenClaw currently only passes `temperature` and `maxTokens` to model providers, ignoring `frequency_penalty` and `presence_penalty` parameters that are critical for
preventing repetitive ... | open | null | false | 4 | [
"enhancement"
] | [] | 2026-03-03T03:48:31Z | 2026-05-05T02:56:45Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | moyao-miao | 95,124,633 | U_kgDOBat8mQ | User | false |
openclaw/openclaw | 4,163,156,027 | I_kwDOQb6kR874JLg7 | 56,854 | https://github.com/openclaw/openclaw/issues/56854 | https://api.github.com/repos/openclaw/openclaw/issues/56854 | [Bug]: Gateway crashes with uncaught exception when Discord health monitor triggers stale-socket restart | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Discord provider crashes the entire gateway with an uncaught exception when the health monitor detects a stale socket and attempts a restart. The abort handler sets maxReconnectAttempts to 0, then the WebSocket close handler... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-29T07:25:23Z | 2026-04-18T04:14:30Z | 2026-04-15T06:40:34Z | NONE | null | 20260418T042035Z | 2026-04-18T04:20:35Z | waqarahmadweb | 34,026,578 | MDQ6VXNlcjM0MDI2NTc4 | User | false |
openclaw/openclaw | 4,194,839,844 | I_kwDOQb6kR876CC0k | 59,702 | https://github.com/openclaw/openclaw/issues/59702 | https://api.github.com/repos/openclaw/openclaw/issues/59702 | sessions_spawn and internal RPC fail with 'gateway closed (1008): pairing required' on Windows Server 2025 | ## Bug Description
sessions_spawn and internal RPC fail with WebSocket close code 1008 'pairing required' on Windows Server 2025 (loopback bind). Main session works fine but all internal agent-to-gateway RPC is blocked.
## Environment
- OpenClaw version: 2026.4.1
- OS: Windows Server 2025 (10.0.26100)
- Node.js: v24... | closed | completed | true | 2 | [] | [] | 2026-04-02T14:04:50Z | 2026-04-29T04:39:49Z | 2026-04-26T09:03:01Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | mortimer22oc-arch | 260,152,630 | U_kgDOD4GdNg | User | false |
openclaw/openclaw | 4,222,754,139 | I_kwDOQb6kR877sh1b | 62,997 | https://github.com/openclaw/openclaw/issues/62997 | https://api.github.com/repos/openclaw/openclaw/issues/62997 | [Bug]: Spawned ACP background task enters zombie state after billing failure — persists across session resets, gateway restarts, and database wipes | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
`<h2 class="text-text-100 mt-3 -mb-1 text-[1.125rem] font-bold">Bug type</h2>
<p class="font-claude-response-body break-words whitespace-normal leading-[1.7]">Behavior bug (incorrect output/state without crash)... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [
"vincentkoc"
] | 2026-04-08T07:10:37Z | 2026-04-28T05:04:45Z | 2026-04-25T14:51:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | WindcentNo | 127,996,034 | U_kgDOB6EQgg | User | false |
openclaw/openclaw | 4,230,933,224 | I_kwDOQb6kR878Luro | 63,654 | https://github.com/openclaw/openclaw/issues/63654 | https://api.github.com/repos/openclaw/openclaw/issues/63654 | Qwen: qwen3.6-plus image understanding blocked on Coding Plan endpoint despite model supporting vision | ## Summary
`qwen3.6-plus` is a multimodal model that supports image understanding, but the bundled `qwen` provider explicitly suppresses it when the configured endpoint is the Qwen Coding Plan (`coding.dashscope.aliyuncs.com`). This prevents the `image` tool and media-understanding pipeline from using `qwen3.6-plus` f... | closed | completed | true | 2 | [] | [] | 2026-04-09T09:03:33Z | 2026-05-01T04:44:04Z | 2026-04-28T09:40:37Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | xijuangu | 125,733,029 | U_kgDOB36IpQ | User | false |
openclaw/openclaw | 4,074,360,095 | I_kwDOQb6kR87y2c0f | 45,645 | https://github.com/openclaw/openclaw/issues/45645 | https://api.github.com/repos/openclaw/openclaw/issues/45645 | [Bug]: doctor/update can leave CLI and gateway on different installs on Linux | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When two OpenClaw installs coexist on Linux (for example an older CLI in `~/.npm-global` and a newer gateway/service install under an nvm-managed Node path), `openclaw doctor` / update flows can leave the shell CLI and the running gateway on ... | closed | completed | true | 4 | [] | [] | 2026-03-14T02:21:26Z | 2026-04-27T04:45:03Z | 2026-04-25T00:39:34Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | Lucenx9 | 185,146,821 | U_kgDOCwkdxQ | User | false |
openclaw/openclaw | 4,058,383,373 | I_kwDOQb6kR87x5gQN | 43,235 | https://github.com/openclaw/openclaw/issues/43235 | https://api.github.com/repos/openclaw/openclaw/issues/43235 | [Feature]:Support per-agent command lanes for multi-agent deployments | ### Summary
In multi-agent deployments (30+ agents, 60+ users), all inbound messages share a single global main lane controlled by agents.defaults.maxConcurrent. When one agent runs a long task (e.g., PPT generation with subagents, multi-step tool calls taking 2-5 minutes), it holds a slot in the shared main lane, blo... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-11T13:54:28Z | 2026-05-04T12:04:22Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | lijihua2017 | 33,613,942 | MDQ6VXNlcjMzNjEzOTQy | User | false |
openclaw/openclaw | 4,213,749,314 | I_kwDOQb6kR877KLZC | 62,102 | https://github.com/openclaw/openclaw/issues/62102 | https://api.github.com/repos/openclaw/openclaw/issues/62102 | Paperclip Gateway adapter sends unsupported 'paperclip' field in agent params |
## Bug Description
When using Paperclip's openclaw_gateway adapter to connect to OpenClaw Gateway, the agent execution fails with:
```
invalid agent params: at root: unexpected property 'paperclip'
```
## Root Cause
The OpenClaw Gateway's `AgentParamsSchema` has `additionalProperties: false` and does not include a... | open | null | false | 3 | [] | [] | 2026-04-06T19:49:31Z | 2026-05-02T05:56:25Z | null | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | schliesing | 170,745,673 | U_kgDOCi1fSQ | User | false |
openclaw/openclaw | 4,025,742,857 | I_kwDOQb6kR87v8_YJ | 35,462 | https://github.com/openclaw/openclaw/issues/35462 | https://api.github.com/repos/openclaw/openclaw/issues/35462 | Reminder-guard: detect follow-up promises (CN + more EN variants) when no scheduler called | # Issue draft: Reminder-guard coverage for follow-up promises (CN + more EN variants)
## Problem
Users often experience: Agent says “稍后同步进度 / 完成后通知 / 到时候告诉你 / I'll let you know when done”, but no message ever arrives.
This is expected given the agent loop is synchronous; however, the system currently has a reminder-p... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T04:50:00Z | 2026-04-24T04:26:37Z | 2026-04-21T04:24:31Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | cathyjay | 114,842,176 | U_kgDOBthaQA | User | false |
openclaw/openclaw | 4,025,736,019 | I_kwDOQb6kR87v89tT | 35,459 | https://github.com/openclaw/openclaw/issues/35459 | https://api.github.com/repos/openclaw/openclaw/issues/35459 | 测试 | 来自openclaw的测试 | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T04:48:05Z | 2026-04-24T04:26:38Z | 2026-04-21T04:24:30Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | shuaigeyuan | 42,385,656 | MDQ6VXNlcjQyMzg1NjU2 | User | false |
openclaw/openclaw | 4,025,728,156 | I_kwDOQb6kR87v87yc | 35,457 | https://github.com/openclaw/openclaw/issues/35457 | https://api.github.com/repos/openclaw/openclaw/issues/35457 | feat(thinking): provider-neutral low-thinking hint for simple turns | ## Problem
For many short/simple turns, OpenClaw may spend more reasoning effort than needed, even when staying on the same model is preferable.
That can increase latency/cost for trivial requests.
## Proposal
Add a lightweight simple-turn heuristic in reply flow:
- If effective think level resolves to `adaptive`
... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T04:45:51Z | 2026-04-24T04:26:39Z | 2026-04-21T04:24:28Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | aronchick | 51,317 | MDQ6VXNlcjUxMzE3 | User | false |
openclaw/openclaw | 4,025,698,664 | I_kwDOQb6kR87v80lo | 35,443 | https://github.com/openclaw/openclaw/issues/35443 | https://api.github.com/repos/openclaw/openclaw/issues/35443 | llm-task plugin tool parameters missing `type` field — breaks llama.cpp backends | ## Description
The `llm-task` plugin tool has two parameters (`input` and `schema`) whose JSON Schema definitions contain only a `description` field without a `type` field. This is technically valid JSON Schema (implies "any type"), but llama.cpp's OpenAI-compatible endpoint rejects it with a 400 error during JSON sch... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T04:37:09Z | 2026-04-24T04:26:40Z | 2026-04-21T04:24:26Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | jmzlx | 665,485 | MDQ6VXNlcjY2NTQ4NQ== | User | false |
openclaw/openclaw | 4,025,681,155 | I_kwDOQb6kR87v8wUD | 35,440 | https://github.com/openclaw/openclaw/issues/35440 | https://api.github.com/repos/openclaw/openclaw/issues/35440 | Feature Request: UI Wizard for Channel Configuration | ## Feature Request: UI Simplificada para Conectar Canais
**Problema:** Usuários precisam editar JSON e rodar comandos bash para conectar Telegram
**Solução:** Wizard no dashboard com:
- [ ] Campo único para token
- [ ] Botão "Conectar"
- [ ] Status visual (🟢 Conectado / 🔴 Desconectado)
- [ ] Auto-descoberta do ID d... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T04:31:27Z | 2026-04-24T04:26:41Z | 2026-04-21T04:24:25Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | escoladetecnologia | 96,396,925 | U_kgDOBb7mfQ | User | false |
openclaw/openclaw | 4,025,670,761 | I_kwDOQb6kR87v8txp | 35,434 | https://github.com/openclaw/openclaw/issues/35434 | https://api.github.com/repos/openclaw/openclaw/issues/35434 | tools.subagents.tools.deny override not applied to subagent sessions | ## Bug
`tools.subagents.tools.deny` config is accepted by the Zod schema but not applied at runtime. Subagents still get the hardcoded default deny list regardless of config overrides.
### Expected behavior
Setting `tools.subagents.tools.deny: ["gateway"]` should override the default `["gateway", "cron"]`, making `c... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T04:28:35Z | 2026-04-24T04:26:41Z | 2026-04-21T04:24:23Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | jacobaross | 16,659,674 | MDQ6VXNlcjE2NjU5Njc0 | User | false |
openclaw/openclaw | 4,025,656,858 | I_kwDOQb6kR87v8qYa | 35,428 | https://github.com/openclaw/openclaw/issues/35428 | https://api.github.com/repos/openclaw/openclaw/issues/35428 | [Bug]: Control UI - Core Files refresh never shows external edits | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In the Control UI “Core Files” panel, the Refresh button only reloads the file list metadata. It never re-fetches the active file’s contents, so any edits made from the CLI (or another editor) are invisible. Saving from the UI after that si... | closed | not_planned | true | 3 | [
"bug",
"stale"
] | [] | 2026-03-05T04:24:57Z | 2026-04-24T04:26:42Z | 2026-04-21T04:24:21Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | trip-null | 198,066,939 | U_kgDOC85C-w | User | false |
openclaw/openclaw | 4,025,630,857 | I_kwDOQb6kR87v8kCJ | 35,414 | https://github.com/openclaw/openclaw/issues/35414 | https://api.github.com/repos/openclaw/openclaw/issues/35414 | Plugin hook: truncateBefore cursor in before_prompt_build result | ## Summary
Allow plugins to drop old messages from the prompt context via a `truncateBefore` timestamp cursor returned from the `before_prompt_build` hook result.
## Motivation
Plugins that manage context — memory systems, conversation summarizers, context compressors — often inject compressed representations of older... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T04:16:58Z | 2026-04-24T04:26:43Z | 2026-04-21T04:24:19Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | unboxed-ai | 253,808,488 | U_kgDODyDPaA | User | false |
openclaw/openclaw | 4,025,612,858 | I_kwDOQb6kR87v8fo6 | 35,405 | https://github.com/openclaw/openclaw/issues/35405 | https://api.github.com/repos/openclaw/openclaw/issues/35405 | **Bug Report: Chrome Extension handshake rejected by gateway (nonce mismatch)** | ### Bug type
Regression (worked before, now fails)
### Summary
**Version:** OpenClaw 2026.3.2 (also reproduced on 2026.3.1)
**OS:** Windows 11, Node 24.14.0
**Install method:** npm global (`npm install -g openclaw`)
**Summary:**
The bundled Chrome extension (`assets/chrome-extension/background.js`) sends a `nonce`... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T04:12:10Z | 2026-04-24T04:26:44Z | 2026-04-21T04:24:18Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | zyang2k | 67,810,238 | MDQ6VXNlcjY3ODEwMjM4 | User | false |
openclaw/openclaw | 4,025,612,230 | I_kwDOQb6kR87v8ffG | 35,404 | https://github.com/openclaw/openclaw/issues/35404 | https://api.github.com/repos/openclaw/openclaw/issues/35404 | Hardening follow-up: tighten subagent post-spawn verification fail-open paths | ## Context
Follow-up to #35248 and PR #35273.
The current fix adds post-spawn verification + bounded retry, which addresses the primary STARTED-without-evidence stall mode.
## Remaining risk
Verification currently fails open when:
1. `sessions.list` throws
2. `sessions.list` response shape is unexpected (no array)
T... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T04:12:01Z | 2026-04-24T04:26:44Z | 2026-04-21T04:24:16Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | ArgadronRey | 183,376,875 | U_kgDOCu4b6w | User | false |
openclaw/openclaw | 4,025,597,366 | I_kwDOQb6kR87v8b22 | 35,397 | https://github.com/openclaw/openclaw/issues/35397 | https://api.github.com/repos/openclaw/openclaw/issues/35397 | [Bug]: OPENCLAW_DOCKER_APT_PACKAGES is not obeyed in setup-podman.sh | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
OPENCLAW_DOCKER_APT_PACKAGES is not obeyed in `setup-podman.sh`
### Steps to reproduce
export OPENCLAW_DOCKER_APT_PACKAGES="some package"
sudo ./setup-podman.sh
### Expected behavior
Installs the requested packages
### Actual behavior
... | closed | not_planned | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T04:08:14Z | 2026-04-24T04:26:45Z | 2026-04-21T04:24:14Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | zydtiger | 43,339,228 | MDQ6VXNlcjQzMzM5MjI4 | User | false |
openclaw/openclaw | 4,025,572,841 | I_kwDOQb6kR87v8V3p | 35,389 | https://github.com/openclaw/openclaw/issues/35389 | https://api.github.com/repos/openclaw/openclaw/issues/35389 | Bug: ACP sessions silently fail (code 5) despite permissionMode approve-all — config loads but turns never execute | ## Environment
- **OpenClaw:** v2026.3.2
- **acpx:** 0.1.15
- **macOS arm64**, Node v24.13.0
- **Agents tested:** Pi (0.55.4), Claude Code (2.1.68), Codex CLI
## Symptom
`sessions_spawn(runtime: "acp")` returns `status: "accepted"` with a valid `childSessionKey` and `runId`, but the task **never executes**. Session ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T04:01:09Z | 2026-04-24T04:26:46Z | 2026-04-21T04:24:13Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | telegraph-it | 127,995,520 | U_kgDOB6EOgA | User | false |
openclaw/openclaw | 4,025,474,286 | I_kwDOQb6kR87v79zu | 35,347 | https://github.com/openclaw/openclaw/issues/35347 | https://api.github.com/repos/openclaw/openclaw/issues/35347 | [Bug]: Frequently "getting Invalid diff: now finding less tool calls!" error on v2026.3.2 | ### Bug type
Regression (worked before, now fails)
### Summary
I'm using the Qwen 3.5 35B model running locally via llama-server. It has been working great for the past couple of days but frequently shows the error "Invalid diff: now finding less tool calls!" on version 2026.3.2. I tried turning off some of the tool... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T03:31:29Z | 2026-04-24T04:26:47Z | 2026-04-21T04:24:11Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | kendy92 | 12,098,762 | MDQ6VXNlcjEyMDk4NzYy | User | false |
openclaw/openclaw | 4,025,426,926 | I_kwDOQb6kR87v7yPu | 35,310 | https://github.com/openclaw/openclaw/issues/35310 | https://api.github.com/repos/openclaw/openclaw/issues/35310 | [Bug] Node notifications.changed wake overrides configured heartbeat.session via forced sessionKey | ## Summary
`heartbeat.session` can be overridden by node notification wake events (`notifications.changed`) because the wake path passes a forced `sessionKey` (`node-<nodeId>`), which then canonicalizes to `agent:main:node-...`.
This causes heartbeat runs to execute in a node-scoped session even when heartbeat session... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-05T03:14:41Z | 2026-04-24T04:26:48Z | 2026-04-21T04:24:09Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | choury | 4,332,542 | MDQ6VXNlcjQzMzI1NDI= | User | false |
openclaw/openclaw | 4,025,422,685 | I_kwDOQb6kR87v7xNd | 35,308 | https://github.com/openclaw/openclaw/issues/35308 | https://api.github.com/repos/openclaw/openclaw/issues/35308 | Bug: deltaBuffer missing cross-turn separator causes text concatenation in streaming UI | ## Summary
When `blockStreaming` is enabled, the agent runner's `deltaBuffer` does not insert any separator between successive API turns. This causes the text from separate tool-call cycles to be concatenated together in the streaming UI, producing garbled output (e.g. pre-tool text runs directly into post-tool text w... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T03:13:05Z | 2026-04-24T04:26:48Z | 2026-04-21T04:24:07Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | matthewangus | 262,665,369 | U_kgDOD6f0mQ | User | false |
openclaw/openclaw | 4,025,419,689 | I_kwDOQb6kR87v7wep | 35,306 | https://github.com/openclaw/openclaw/issues/35306 | https://api.github.com/repos/openclaw/openclaw/issues/35306 | [Bug]: screen_record is blocked and system.which fails | ### Bug type
Regression (worked before, now fails)
### Summary
Environment
• Gateway: OpenClaw 2026.3.2 (Linux)
• Node app: OpenClaw 2026.3.2
• Node platform: macOS 26.2.0
• Node: MacBook Air
Issue
nodes describe reports:
• caps: includes screen
• commands: includes system.run, system.which
• permissions.screenRe... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T03:12:05Z | 2026-04-24T04:26:49Z | 2026-04-21T04:24:05Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | arlenops | 175,078,360 | U_kgDOCm972A | User | false |
openclaw/openclaw | 4,025,415,154 | I_kwDOQb6kR87v7vXy | 35,302 | https://github.com/openclaw/openclaw/issues/35302 | https://api.github.com/repos/openclaw/openclaw/issues/35302 | Feature request: pre-send / post-send response hooks | ## Use Case
We're building a self-improvement loop where an AI agent's outbound responses get checked (regex + cheap model) for quality issues like permission-seeking, option parades, and tone mismatches.
Currently we're doing this by tailing the session JSONL files and parsing assistant messages — it works but it's ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T03:10:26Z | 2026-04-24T04:26:50Z | 2026-04-21T04:24:04Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | ajb181 | 150,086,301 | U_kgDOCPIinQ | User | false |
openclaw/openclaw | 4,025,408,901 | I_kwDOQb6kR87v7t2F | 35,297 | https://github.com/openclaw/openclaw/issues/35297 | https://api.github.com/repos/openclaw/openclaw/issues/35297 | [Bug]: Control UI: Tools toggles don’t persist / Save disabled on Windows (v2026.3.2) | ### Bug type
Regression (worked before, now fails)
### Summary
In Control UI, tool toggles can be changed visually, but Save is disabled or changes do not persist. “Enable all/Disable all” and selective toggling are unreliable.
### Steps to reproduce
Steps to reproduce:
Open Dashboard (http://127.0.0.1:18789/)
Go... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T03:08:18Z | 2026-04-24T04:26:51Z | 2026-04-21T04:24:02Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | Openclawaibuddy | 265,695,320 | U_kgDOD9YwWA | User | false |
openclaw/openclaw | 4,025,398,916 | I_kwDOQb6kR87v7raE | 35,289 | https://github.com/openclaw/openclaw/issues/35289 | https://api.github.com/repos/openclaw/openclaw/issues/35289 | Slug-gen model override (avoid burning primary model RPM on slug generation) | ## Problem
The session-memory slug generator (`slug-gen`) always uses the primary model configured for the session. On rate-limited providers, this compounds the problem:
1. A tiny slug generation call (just generating a short filename) consumes an API request from the same RPM/RPD budget as the main session work.
2.... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T03:04:23Z | 2026-04-24T04:26:52Z | 2026-04-21T04:24:00Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | salvador-clawi | 261,618,414 | U_kgDOD5f67g | User | false |
openclaw/openclaw | 4,025,398,296 | I_kwDOQb6kR87v7rQY | 35,288 | https://github.com/openclaw/openclaw/issues/35288 | https://api.github.com/repos/openclaw/openclaw/issues/35288 | Built-in rate limit notifications | ## Problem
When provider rate limits are hit, there is no user-facing alerting. Users discover the problem only by reading raw gateway logs — which most users never check. Sessions silently stall or retry without any indication.
On 2026-03-05, a thundering-herd rate-limit storm went unnoticed for an extended period b... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T03:04:08Z | 2026-04-24T04:26:52Z | 2026-04-21T04:23:58Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | salvador-clawi | 261,618,414 | U_kgDOD5f67g | User | false |
openclaw/openclaw | 4,025,397,743 | I_kwDOQb6kR87v7rHv | 35,287 | https://github.com/openclaw/openclaw/issues/35287 | https://api.github.com/repos/openclaw/openclaw/issues/35287 | Configurable provider cooldown / fail-fast mode on 429s | ## Problem
When a provider returns HTTP 429 (rate limit), OpenClaw enters a hardcoded exponential backoff schedule (1m → 5m → 25m → 1h) with no user visibility. The user's session silently stalls with no indication of what's happening or how long the wait will be.
On 2026-03-05, this created a thundering-herd scenari... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T03:03:56Z | 2026-04-24T04:26:53Z | 2026-04-21T04:23:56Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | salvador-clawi | 261,618,414 | U_kgDOD5f67g | User | false |
openclaw/openclaw | 4,025,396,931 | I_kwDOQb6kR87v7q7D | 35,285 | https://github.com/openclaw/openclaw/issues/35285 | https://api.github.com/repos/openclaw/openclaw/issues/35285 | Telegram skill commands registered multiple times (_2, _3 suffixes) | ## Description
Skill commands are being registered with Telegram bot menu multiple times, creating duplicates with `_2` and `_3` suffixes.
## Example
Instead of:
- `/weather`
- `/gog`
- `/github`
The bot menu shows:
- `/weather`, `/weather_2`, `/weather_3`
- `/gog`, `/gog_2`, `/gog_3`
- `/github`, `/github_2`, `/gith... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T03:03:36Z | 2026-04-24T04:26:54Z | 2026-04-21T04:23:55Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | egantz | 4,982,233 | MDQ6VXNlcjQ5ODIyMzM= | User | false |
openclaw/openclaw | 4,025,384,258 | I_kwDOQb6kR87v7n1C | 35,283 | https://github.com/openclaw/openclaw/issues/35283 | https://api.github.com/repos/openclaw/openclaw/issues/35283 | [Bug] Gateway restart causes Telegram group sessions to reset instead of restoring from sessions.json | ### Summary
After restarting the OpenClaw gateway, Telegram group-bound agents create fresh sessions instead of restoring existing sessions from `sessions.json`. The session key mapping exists in `sessions.json`, but the gateway creates a new session with `action:"new"` rather than loading the previous session.
### En... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:59:35Z | 2026-04-24T04:26:55Z | 2026-04-21T04:23:53Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | crazyn2 | 45,783,572 | MDQ6VXNlcjQ1NzgzNTcy | User | false |
openclaw/openclaw | 4,025,364,517 | I_kwDOQb6kR87v7jAl | 35,277 | https://github.com/openclaw/openclaw/issues/35277 | https://api.github.com/repos/openclaw/openclaw/issues/35277 | openclaw-tui: Messages displaying twice in terminal UI | **Description:** Messages appear duplicated in openclaw-tui window, but display correctly in Telegram (single message).\n\n**Steps to reproduce:**\n1. Use openclaw-tui interface\n2. Send/receive messages\n\n**Expected behavior:** Single message displayed\n\n**Actual behavior:** Messages rendered twice in TUI\n\n**Envir... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:53:26Z | 2026-04-24T04:26:55Z | 2026-04-21T04:23:51Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | ziscross-gh | 121,753,307 | U_kgDOB0HO2w | User | false |
openclaw/openclaw | 4,025,353,792 | I_kwDOQb6kR87v7gZA | 35,275 | https://github.com/openclaw/openclaw/issues/35275 | https://api.github.com/repos/openclaw/openclaw/issues/35275 | [Bug]: feishu：receive events or callbacks through persistent connection only available in self-build & Feishu app | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
问答没有回复
### Steps to reproduce
飞书开发者后台配置：
<img width="1770" height="795" alt="Image" src="https://github.com/user-attachments/assets/d1e32790-d9c4-48ad-abc8-7ec0a10a582a" />
channel：
<img width="833" height="685" alt="Image" src="https://... | closed | not_planned | true | 2 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-05T02:50:18Z | 2026-04-24T04:26:56Z | 2026-04-21T04:23:50Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | ms20hj | 23,376,892 | MDQ6VXNlcjIzMzc2ODky | User | false |
openclaw/openclaw | 4,025,347,094 | I_kwDOQb6kR87v7ewW | 35,272 | https://github.com/openclaw/openclaw/issues/35272 | https://api.github.com/repos/openclaw/openclaw/issues/35272 | [BUG] Sub-agents cannot access shared skills from ~/.openclaw/skills despite official documentation stating they should be visible to all agents | ## Describe the bug
In a multi-agent setup, sub-agents (david, chris, max, merry, iris) are unable to access skills installed in the global shared skills directory (`~/.openclaw/skills/`), even though the official documentation explicitly states that shared skills should be "visible to all agents on the same machine."... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:48:06Z | 2026-04-24T04:26:57Z | 2026-04-21T04:23:48Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | OliverVin | 20,222,136 | MDQ6VXNlcjIwMjIyMTM2 | User | false |
openclaw/openclaw | 4,025,322,022 | I_kwDOQb6kR87v7Yom | 35,257 | https://github.com/openclaw/openclaw/issues/35257 | https://api.github.com/repos/openclaw/openclaw/issues/35257 | Gateway crash: Uncaught TypeError in undici TLS setSession during sub-agent failover storm | ## Bug Description
Gateway crashes with an uncaught exception when multiple sub-agents trigger rapid provider failover, causing a TLS connection storm in undici.
## Error
```
Uncaught exception: TypeError: Cannot read properties of null (reading 'setSession')
at TLSSocket.setSession (node:_tls_wrap:1132:16)
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T02:37:59Z | 2026-04-24T04:26:58Z | 2026-04-21T04:23:42Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | Simon-Copilot-Studio | 199,782,277 | U_kgDOC-hvhQ | User | false |
openclaw/openclaw | 4,025,306,726 | I_kwDOQb6kR87v7U5m | 35,254 | https://github.com/openclaw/openclaw/issues/35254 | https://api.github.com/repos/openclaw/openclaw/issues/35254 | [Bug]: TUI messages routed to Discord DM instead of TUI | ### Bug type
Regression (worked before, now fails)
### Summary
## Bug: TUI messages routed to Discord DM instead of TUI
### Descr... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T02:32:31Z | 2026-04-24T04:26:58Z | 2026-04-21T04:23:41Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | lefeiren | 115,707,189 | U_kgDOBuWNNQ | User | false |
openclaw/openclaw | 4,025,301,886 | I_kwDOQb6kR87v7Tt- | 35,249 | https://github.com/openclaw/openclaw/issues/35249 | https://api.github.com/repos/openclaw/openclaw/issues/35249 | [Bug]: After update to the latest version of OpenClaw (Cannot enable Browser - Google Chrome) | ### Bug type
Regression (worked before, now fails)
### Summary
Before the Update I was able to use OpenClaw , I enabled headless browser and it browsed and did some tasks for me but after I update when I click on Agents - Tools and enable :
browser core
Control web browser
canvas core
Control canvases
I cannot sa... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T02:31:09Z | 2026-04-24T04:26:59Z | 2026-04-21T04:23:39Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | Marto88 | 79,672,606 | MDQ6VXNlcjc5NjcyNjA2 | User | false |
openclaw/openclaw | 4,025,300,752 | I_kwDOQb6kR87v7TcQ | 35,248 | https://github.com/openclaw/openclaw/issues/35248 | https://api.github.com/repos/openclaw/openclaw/issues/35248 | Auto-next chain can commit STARTED without verifiable spawned session, causing phantom-active stall | ## Summary
Auto-next orchestration can enter a false-STARTED state and then stall with phantom-active-session gating.
## Observed behavior
In an automation setup with:
- one cron job as **single launch authority**
- one cron job as **alert-only watchdog**
- bridge/SSOT state file for completion->spawn handoff
we inte... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:30:51Z | 2026-04-24T04:27:00Z | 2026-04-21T04:23:37Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | ArgadronRey | 183,376,875 | U_kgDOCu4b6w | User | false |
openclaw/openclaw | 4,025,284,676 | I_kwDOQb6kR87v7PhE | 35,242 | https://github.com/openclaw/openclaw/issues/35242 | https://api.github.com/repos/openclaw/openclaw/issues/35242 | [Feature]: Configurable context removal on Slack message_deleted events | ### Problem
OpenClaw currently has no option to control what happens to agent context when a Slack message is deleted. The only behavior is a one-time system event (`"Slack message deleted in #channel."`) with no effect on session history, and no way to configure otherwise.
The deletion event carries the original mes... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:25:30Z | 2026-04-24T04:27:01Z | 2026-04-21T04:23:36Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | juppytt | 46,583,939 | MDQ6VXNlcjQ2NTgzOTM5 | User | false |
openclaw/openclaw | 4,025,268,446 | I_kwDOQb6kR87v7Lje | 35,236 | https://github.com/openclaw/openclaw/issues/35236 | https://api.github.com/repos/openclaw/openclaw/issues/35236 | [Feature Request] Implement MCP Tool Search with Lazy Loading for Context Window Optimization | ## 📋 Summary
Implement a Claude Code-like **MCP Tool Search with Lazy Loading** mechanism to optimize context window usage when multiple MCP servers are connected.
## 🎯 Motivation
OpenClaw supports multiple AI providers (z.ai, OpenRouter, Groq, etc.), and some models have smaller context windows. When many MCP ser... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:19:44Z | 2026-04-24T04:27:02Z | 2026-04-21T04:23:32Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | downwind7clawd-ctrl | 258,506,560 | U_kgDOD2h_QA | User | false |
openclaw/openclaw | 4,025,247,602 | I_kwDOQb6kR87v7Gdy | 35,227 | https://github.com/openclaw/openclaw/issues/35227 | https://api.github.com/repos/openclaw/openclaw/issues/35227 | [Bug]: The large model is unresponsive and does not reply. In addition, the messages I entered are not saved. | ### Bug type
Regression (worked before, now fails)
### Summary
[openclaw-bug-2026-03-05_09.58.10.mkv.txt](https://github.com/user-attachments/files/25755247/openclaw-bug-2026-03-05_09.58.10.mkv.txt)
please check the screen recording file for the detail of the bug.The large model is unresponsive and does not reply. I... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T02:12:41Z | 2026-04-24T04:27:02Z | 2026-04-21T04:23:28Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | RayYe586 | 20,855,804 | MDQ6VXNlcjIwODU1ODA0 | User | false |
openclaw/openclaw | 4,025,224,035 | I_kwDOQb6kR87v7Atj | 35,220 | https://github.com/openclaw/openclaw/issues/35220 | https://api.github.com/repos/openclaw/openclaw/issues/35220 | [Bug]: Codex Responses API streaming server_error does not trigger model fallback (variant of #24378) | ## Summary
When the OpenAI Codex Responses API returns a `server_error` within a streaming response, the session terminates with `stopReason: "error"` but **no fallback model is attempted**, despite a fully configured fallback chain.
This is a variant of #24378 — same root cause (assistant-message errors bypassing `r... | closed | not_planned | true | 16 | [
"stale"
] | [] | 2026-03-05T02:05:40Z | 2026-04-24T04:27:03Z | 2026-04-21T04:23:27Z | CONTRIBUTOR | null | 20260424T060021Z | 2026-04-24T06:00:21Z | hanamizuki | 741,302 | MDQ6VXNlcjc0MTMwMg== | User | false |
openclaw/openclaw | 4,025,222,886 | I_kwDOQb6kR87v7Abm | 35,219 | https://github.com/openclaw/openclaw/issues/35219 | https://api.github.com/repos/openclaw/openclaw/issues/35219 | Discord: ignoreOtherMentions documented but rejected by config schema (2026.3.2) | ## Summary
`ignoreOtherMentions` is documented for Discord guild policy, but config validation rejects it as an unrecognized key.
## Environment
- OpenClaw CLI: `2026.3.2`
- Gateway app: `2026.3.2`
- OS: Linux arm64
- Node: `v22.22.0`
## Expected
`ignoreOtherMentions` should be accepted in Discord guild policy config... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:05:21Z | 2026-04-24T04:27:04Z | 2026-04-21T04:23:25Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | openvheng-bot | 264,998,855 | U_kgDOD8uPxw | User | false |
openclaw/openclaw | 4,025,206,394 | I_kwDOQb6kR87v68Z6 | 35,215 | https://github.com/openclaw/openclaw/issues/35215 | https://api.github.com/repos/openclaw/openclaw/issues/35215 | Session orphaned when heartbeat/cron hits a session with a different model | ## Session orphaned when heartbeat/cron hits a session with a different model
### Summary
When a heartbeat or cron job targets a session that was created with a different model, the gateway creates a new session and remaps the session key — but the previous session file is not backed up as `.reset`. The old session b... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T02:00:18Z | 2026-04-24T04:27:05Z | 2026-04-21T04:23:23Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | nova-oc-bot | 262,968,592 | U_kgDOD6yVEA | User | false |
openclaw/openclaw | 4,025,182,541 | I_kwDOQb6kR87v62lN | 35,202 | https://github.com/openclaw/openclaw/issues/35202 | https://api.github.com/repos/openclaw/openclaw/issues/35202 | [Bug]: feishu配置长链接失败 | ### Bug type
Regression (worked before, now fails)
### Summary
openclaw配置完feishu channel，登陆飞书后台保存长链接失败，日志显示：
[ws]', 'protocol mismatch'
### Steps to reproduce
使用2026.3.2或者3.1都不行，回退到2.25就可以了
### Expected behavior
长链接保存成功
### Actual behavior
未检测到应用长链接
### OpenClaw version
2026.3.2
### Operating system
macOS ... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T01:52:44Z | 2026-04-25T04:19:00Z | 2026-04-22T04:31:14Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | DisguiseSmart | 83,894,937 | MDQ6VXNlcjgzODk0OTM3 | User | false |
openclaw/openclaw | 4,025,157,855 | I_kwDOQb6kR87v6wjf | 35,192 | https://github.com/openclaw/openclaw/issues/35192 | https://api.github.com/repos/openclaw/openclaw/issues/35192 | Maintenance: Gateway 122 commits behind main (srv1325349) | ## Overnight Infrastructure Audit — 2026-03-05
### Issue
Gateway on srv1325349 is **122 commits behind** origin/main. Current npm latest: `2026.3.2`
### Server
- **Host:** srv1325349 (Hostinger)
- **Uptime:** 11 days
- **Status:** Stable (no errors in logs)
### Recommendation
Schedule maintenance window to run:
```b... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T01:43:00Z | 2026-04-25T04:19:01Z | 2026-04-22T04:31:10Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | TerminalGravity | 15,236,602 | MDQ6VXNlcjE1MjM2NjAy | User | false |
openclaw/openclaw | 4,025,129,912 | I_kwDOQb6kR87v6pu4 | 35,181 | https://github.com/openclaw/openclaw/issues/35181 | https://api.github.com/repos/openclaw/openclaw/issues/35181 | [Bug]: [node-llama-cpp] A prebuilt binary was not found, falling back to using no GPU | ### Bug type
Regression (worked before, now fails)
### Summary
Win11
node v22.15.0
npm 11.11.0
When i exec "npm i -g openclaw@latest",
The total error is :
```bash
PS C:\WINDOWS\system32> npm cache clean --force
npm warn using --force Recommended protections disabled.
PS C:\WINDOWS\system32> npm i -g openclaw@late... | closed | not_planned | true | 8 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-05T01:33:46Z | 2026-04-25T04:19:02Z | 2026-04-22T04:31:06Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | SenKnight | 15,954,615 | MDQ6VXNlcjE1OTU0NjE1 | User | false |
openclaw/openclaw | 4,025,127,721 | I_kwDOQb6kR87v6pMp | 35,179 | https://github.com/openclaw/openclaw/issues/35179 | https://api.github.com/repos/openclaw/openclaw/issues/35179 | Telegram binding 路由错误：使用 sender_id 而非 bot_account_id 匹配导致消息路由到错误的 agent | ## 问题描述
配置了多个 Telegram bot 账号的 binding 规则，但消息被路由到错误的 agent。
## 配置
```json
{
"bindings": [
{
"agentId": "claw-a",
"match": {
"channel": "telegram",
"accountId": "<BOT_ACCOUNT_ID_A>"
}
},
{
"agentId": "main",
"match": {
"channel": "telegram",
... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T01:33:05Z | 2026-04-25T04:19:03Z | 2026-04-22T04:31:02Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | andrewcapture | 5,107,933 | MDQ6VXNlcjUxMDc5MzM= | User | false |
openclaw/openclaw | 4,025,108,887 | I_kwDOQb6kR87v6kmX | 35,171 | https://github.com/openclaw/openclaw/issues/35171 | https://api.github.com/repos/openclaw/openclaw/issues/35171 | Dashboard UI 版本显示不一致 (v2026.3.2 npm 包显示 v2026.3.1) | ## 问题描述
全局安装 OpenClaw v2026.3.2 后，CLI 和 npm 包版本正确显示为 2026.3.2，但 Dashboard 控制界面右上角版本显示仍为 2026.3.1。
## 环境信息
- **OpenClaw 版本**: 2026.3.2
- **操作系统**: Windows 10 (26200)
- **Node.js**: v24.13.1
- **安装方式**: \
pm install -g openclaw\`n
## 复现步骤
1. \
pm uninstall -g openclaw\`n2. \
pm install -g openclaw\`n3. \openclaw gate... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T01:26:31Z | 2026-04-25T04:19:03Z | 2026-04-22T04:31:00Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | def177 | 139,129,111 | U_kgDOCErxFw | User | false |
openclaw/openclaw | 4,025,105,418 | I_kwDOQb6kR87v6jwK | 35,166 | https://github.com/openclaw/openclaw/issues/35166 | https://api.github.com/repos/openclaw/openclaw/issues/35166 | Browser Tool Bug: targetUrl required Error Blocks Testing (Revised) | # OpenClaw Browser Tool Bug Report
**Submitted**: March 4, 2026
**Issue**: Persistent 'targetUrl required' error with browser tool actions (`navigate`, `evaluate`, etc.)
**Description**: When attempting to use the browser tool with actions like `navigate` or `evaluate` after successfully opening a tab with `open` (p... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T01:25:23Z | 2026-04-25T04:19:04Z | 2026-04-22T04:30:56Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | joesinvestments | 257,641,472 | U_kgDOD1tMAA | User | false |
openclaw/openclaw | 4,025,098,924 | I_kwDOQb6kR87v6iKs | 35,160 | https://github.com/openclaw/openclaw/issues/35160 | https://api.github.com/repos/openclaw/openclaw/issues/35160 | fix: OpenAI/Codex server_error not recognized as retryable — model fallback skipped | ## Bug
When OpenAI Codex returns a transient 500 error, the model fallback logic does not trigger. The agent surfaces the raw error to the user instead of falling back to configured alternative models.
## Root Cause
`isJsonApiInternalServerError()` in `src/agents/pi-embedded-helpers/errors.ts` only matches Anthropic... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-05T01:23:10Z | 2026-04-25T04:19:05Z | 2026-04-22T04:30:55Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | ToneLoke | 10,687,151 | MDQ6VXNlcjEwNjg3MTUx | User | false |
openclaw/openclaw | 4,025,097,205 | I_kwDOQb6kR87v6hv1 | 35,157 | https://github.com/openclaw/openclaw/issues/35157 | https://api.github.com/repos/openclaw/openclaw/issues/35157 | Browser Tool Bug: targetUrl required Error Blocks Testing | # OpenClaw Browser Tool Bug Report
**Submitted**: March 4, 2026
**Issue**: Persistent 'targetUrl required' error with browser tool actions (`navigate`, `evaluate`, etc.)
**Description**: When attempting to use the browser tool with actions like `navigate` or `evaluate` after successfully opening a tab with `open` (p... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T01:22:30Z | 2026-04-25T04:19:06Z | 2026-04-22T04:30:53Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | joesinvestments | 257,641,472 | U_kgDOD1tMAA | User | false |
openclaw/openclaw | 4,025,035,319 | I_kwDOQb6kR87v6So3 | 35,134 | https://github.com/openclaw/openclaw/issues/35134 | https://api.github.com/repos/openclaw/openclaw/issues/35134 | [Feature]: skill-creator — add eval-driven description improvement workflow | ## Summary
Let skill authors test their skill descriptions against sample prompts, measure trigger accuracy, and iteratively refine descriptions based on automated feedback.
## Problem to solve
`skill-creator` validates skill format but not quality. There's no way to answer "does this description actually trigger on... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T01:02:54Z | 2026-04-25T04:19:06Z | 2026-04-22T04:30:38Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | iml885203 | 17,023,091 | MDQ6VXNlcjE3MDIzMDkx | User | false |
openclaw/openclaw | 4,025,028,702 | I_kwDOQb6kR87v6RBe | 35,132 | https://github.com/openclaw/openclaw/issues/35132 | https://api.github.com/repos/openclaw/openclaw/issues/35132 | [Feature]: Alice&Bot channel extension | ## Summary
Add Alice&Bot as a channel extension, so OpenClaw agents can communicate through it alongside Telegram, Signal, Matrix, etc.
## Problem to solve
Right now if you want your OpenClaw agent to talk to users over chat, you're constrained to the big incumbents. They all come with friction: phone number require... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T01:01:02Z | 2026-04-23T04:22:42Z | 2026-04-20T04:31:07Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | uriva | 2,490,655 | MDQ6VXNlcjI0OTA2NTU= | User | false |
openclaw/openclaw | 4,025,021,037 | I_kwDOQb6kR87v6PJt | 35,129 | https://github.com/openclaw/openclaw/issues/35129 | https://api.github.com/repos/openclaw/openclaw/issues/35129 | openclaw doctor: detect duplicate installations fighting over gateway port | ## Problem
On a production server, two openclaw installations were present simultaneously:
1. A **system-level** npm install at `/usr/lib/node_modules/openclaw` (old v2026.2.25, installed via `sudo npm install -g`)
2. A **user-level** npm install at `~/.npm-global/lib/node_modules/openclaw` (v2026.3.2, the correct on... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T00:58:25Z | 2026-04-23T04:22:42Z | 2026-04-20T04:31:06Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | aronchick | 51,317 | MDQ6VXNlcjUxMzE3 | User | false |
openclaw/openclaw | 4,025,017,200 | I_kwDOQb6kR87v6ONw | 35,127 | https://github.com/openclaw/openclaw/issues/35127 | https://api.github.com/repos/openclaw/openclaw/issues/35127 | [Feature] Keet (Holepunch) channel plugin for P2P private messaging | ## Summary
Add a native **Keet** channel plugin for OpenClaw, enabling truly peer-to-peer, serverless messaging between users and their agents.
## What is Keet?
[Keet](https://keet.io) is a P2P chat app built on the [Holepunch / Hypercore Protocol](https://holepunch.to). Unlike relay-based protocols (including NOSTR... | closed | not_planned | true | 4 | [] | [] | 2026-03-05T00:56:57Z | 2026-04-27T05:03:15Z | 2026-04-24T19:22:39Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | LoknLod | 124,176,721 | U_kgDOB2bJUQ | User | false |
openclaw/openclaw | 4,024,975,293 | I_kwDOQb6kR87v6D-9 | 35,113 | https://github.com/openclaw/openclaw/issues/35113 | https://api.github.com/repos/openclaw/openclaw/issues/35113 | [Bug]: Gateway crash-loops when gateway.tailscale.mode=serve but gateway.bind != loopback | ## Environment
- **OpenClaw version:** 2026.2.25
- **OS:** Ubuntu 24.04 (DigitalOcean)
- **Node:** 25.6.0
- **Systemd unit:** user-level `openclaw-gateway.service` with `Restart=always`
## Description
When `gateway.tailscale.mode` is set to `"serve"` or `"funnel"` and `gateway.bind` is set to anything other than `"l... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T00:41:39Z | 2026-04-23T04:22:43Z | 2026-04-20T04:31:02Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | aronchick | 51,317 | MDQ6VXNlcjUxMzE3 | User | false |
openclaw/openclaw | 4,024,971,445 | I_kwDOQb6kR87v6DC1 | 35,110 | https://github.com/openclaw/openclaw/issues/35110 | https://api.github.com/repos/openclaw/openclaw/issues/35110 | [Bug]: skill-creator quick_validate.py silently accepts empty name and description | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
`quick_validate.py` passes validation for skills with an empty `name` or `description` field, returning `Skill is valid!` instead of an error. This allows broken skills to be packaged and installed without any warning.
## Steps to reproduce
... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-05T00:40:08Z | 2026-04-23T04:22:44Z | 2026-04-20T04:31:00Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | iml885203 | 17,023,091 | MDQ6VXNlcjE3MDIzMDkx | User | false |
openclaw/openclaw | 4,024,960,419 | I_kwDOQb6kR87v6AWj | 35,107 | https://github.com/openclaw/openclaw/issues/35107 | https://api.github.com/repos/openclaw/openclaw/issues/35107 | /stop command — halt all in-flight tool calls and subagents without killing the session | ## Problem
When the agent goes on a roll (chaining tool calls, spawning subagents), there's no way to hit the brakes without nuking the entire session with `/new`.
Typing "stop" in chat depends on the agent reading it between tool calls — unreliable. Subagents have no kill switch from the chat surface at all.
## Pro... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T00:36:31Z | 2026-04-23T04:22:45Z | 2026-04-20T04:30:58Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | marklights54-byte | 264,699,524 | U_kgDOD8b-hA | User | false |
openclaw/openclaw | 4,024,902,571 | I_kwDOQb6kR87v5yOr | 35,092 | https://github.com/openclaw/openclaw/issues/35092 | https://api.github.com/repos/openclaw/openclaw/issues/35092 | Bug: /new does not flush queued messages, delivering stale messages to fresh session | ## Description
When `/new` (or `/reset`) is issued while messages are queued (agent busy), the queued messages survive the session reset and are delivered as the first input to the new session. This causes the agent to respond to questions from the previous conversation context, appearing to hallucinate.
## Steps to ... | closed | not_planned | true | 5 | [
"stale"
] | [] | 2026-03-05T00:19:43Z | 2026-04-23T04:22:45Z | 2026-04-20T04:30:55Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | sttravis | 265,517,686 | U_kgDOD9N6dg | User | false |
openclaw/openclaw | 4,024,869,037 | I_kwDOQb6kR87v5qCt | 35,083 | https://github.com/openclaw/openclaw/issues/35083 | https://api.github.com/repos/openclaw/openclaw/issues/35083 | hooks/agent: keep summaries in requested session and use hook lane tag | ## Summary
`/hooks/agent` runs can enqueue fallback completion/error summaries into the main session instead of the requested target session. In DM lanes this can make announce/pickup behavior look inconsistent (messages appear in the wrong lane, then may be treated as housekeeping).
## Actual behavior
- Hook-triggere... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-05T00:09:53Z | 2026-04-23T04:22:46Z | 2026-04-20T04:30:54Z | CONTRIBUTOR | null | 20260423T060020Z | 2026-04-23T06:00:20Z | AytuncYildizli | 47,717,026 | MDQ6VXNlcjQ3NzE3MDI2 | User | false |
openclaw/openclaw | 4,024,830,505 | I_kwDOQb6kR87v5gop | 35,078 | https://github.com/openclaw/openclaw/issues/35078 | https://api.github.com/repos/openclaw/openclaw/issues/35078 | [Bug]: Plugin native bindings (node_sqlite3) fail to load under Jiti loader, breaking mem0 OSS mode recall/capture entirely | ### Bug type
Regression (worked before, now fails)
### Summary
When using `@mem0/openclaw-mem0` in `open-source` mode, every `recall` and `capture` call
fails at runtime with a native binding resolution error. The plugin registers and initializes
successfully, but is completely non-functional.
### Steps to reprod... | closed | not_planned | true | 6 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T23:59:49Z | 2026-04-23T04:22:47Z | 2026-04-20T04:30:52Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | vincentrcl000 | 223,226,791 | U_kgDODU4rpw | User | false |
openclaw/openclaw | 4,024,802,666 | I_kwDOQb6kR87v5Z1q | 35,072 | https://github.com/openclaw/openclaw/issues/35072 | https://api.github.com/repos/openclaw/openclaw/issues/35072 | [Bug]: iMessage channel health monitor restarts provider every 30 minutes during normal idle periods | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The channel health monitor uses a fixed 30-minute staleEventThresholdMs to detect dead sockets, then force-restarts the provider. This threshold works for high-traffic channels (Slack/Discord) but iMessage can be legitimately quiet for hour... | closed | completed | true | 6 | [
"bug",
"bug:behavior"
] | [] | 2026-03-04T23:52:46Z | 2026-04-24T04:25:07Z | 2026-04-22T02:33:59Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | mira-lgtm | 265,668,008 | U_kgDOD9XFqA | User | false |
openclaw/openclaw | 4,024,727,283 | I_kwDOQb6kR87v5Hbz | 35,063 | https://github.com/openclaw/openclaw/issues/35063 | https://api.github.com/repos/openclaw/openclaw/issues/35063 | [Bug]: web_search Grok parser crashes on malformed output entries | ### Bug type
Crash (process/app exits or throws)
### Summary
`extractGrokContent` crashes when xAI/Grok Responses payloads contain malformed entries in `output` or `output[].content` (for example `null`).
### Steps to reproduce
1. Call `extractGrokContent` with malformed runtime payload:
```ts
extractGrokContent({... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T23:32:31Z | 2026-04-23T04:22:48Z | 2026-04-20T04:30:49Z | CONTRIBUTOR | null | 20260423T060020Z | 2026-04-23T06:00:20Z | bmendonca3 | 208,517,100 | U_kgDODG237A | User | false |
openclaw/openclaw | 4,024,717,170 | I_kwDOQb6kR87v5E9y | 35,058 | https://github.com/openclaw/openclaw/issues/35058 | https://api.github.com/repos/openclaw/openclaw/issues/35058 | Feature request: Route sub-agent completion announcements to specific channel/session | ## Problem
When using `sessions_spawn` with `mode: "run"`, completions auto-announce to the parent session (the user's DM). This creates noise when running background tasks like QA reviews or engagement scans that should post results to a different channel (e.g., Discord or a separate Telegram group).
## Current beha... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T23:29:08Z | 2026-04-23T04:22:49Z | 2026-04-20T04:30:47Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | drinkwithrapha | 261,669,649 | U_kgDOD5jDEQ | User | false |
openclaw/openclaw | 4,024,699,930 | I_kwDOQb6kR87v5Awa | 35,053 | https://github.com/openclaw/openclaw/issues/35053 | https://api.github.com/repos/openclaw/openclaw/issues/35053 | Feature: Tool Result Middleware / Output Compression Hook | ## Summary
Add a middleware layer that intercepts tool results **before** they are added to the LLM context, allowing post-processors to compress, truncate, or transform outputs.
## Motivation
Tools like [rtk (Rust Token Killer)](https://github.com/rtk-ai/rtk) can compress `exec`, `read_file`, and other tool outputs... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T23:23:23Z | 2026-04-23T04:22:49Z | 2026-04-20T04:30:46Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | erscoder | 9,429,379 | MDQ6VXNlcjk0MjkzNzk= | User | false |
openclaw/openclaw | 4,024,690,956 | I_kwDOQb6kR87v4-kM | 35,051 | https://github.com/openclaw/openclaw/issues/35051 | https://api.github.com/repos/openclaw/openclaw/issues/35051 | [Bug]: openai-ws input conversion crashes on malformed content entries | ### Bug type
Crash (process/app exits or hangs)
### Summary
`convertMessagesToInputItems` can throw when user/assistant content arrays contain malformed entries (e.g. `null`) because `contentToOpenAIParts` and assistant-content iteration dereference `part.type`/`block.type` without object guards.
### Steps to repro... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T23:20:27Z | 2026-04-23T04:22:50Z | 2026-04-20T04:30:44Z | CONTRIBUTOR | null | 20260423T060020Z | 2026-04-23T06:00:20Z | bmendonca3 | 208,517,100 | U_kgDODG237A | User | false |
openclaw/openclaw | 4,024,681,863 | I_kwDOQb6kR87v48WH | 35,045 | https://github.com/openclaw/openclaw/issues/35045 | https://api.github.com/repos/openclaw/openclaw/issues/35045 | [Bug]: openai-ws response parser crashes on malformed message content entries | ### Bug type
Crash (process/app exits or hangs)
### Summary
`buildAssistantMessageFromResponse` can throw when OpenAI WS message content contains malformed entries (for example `null`) because it dereferences `part.type` without object guards.
### Steps to reproduce
1. Build a WS response object with `output: [{ t... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T23:17:30Z | 2026-04-23T04:22:51Z | 2026-04-20T04:30:42Z | CONTRIBUTOR | null | 20260423T060020Z | 2026-04-23T06:00:20Z | bmendonca3 | 208,517,100 | U_kgDODG237A | User | false |
openclaw/openclaw | 4,024,672,294 | I_kwDOQb6kR87v46Am | 35,042 | https://github.com/openclaw/openclaw/issues/35042 | https://api.github.com/repos/openclaw/openclaw/issues/35042 | [Feature Proposal] Eval Framework for skill-creator | ## Background
Anthropic recently published a blog post on [improving skill-creator](https://claude.com/blog/improving-skill-creator-test-measure-and-refine-agent-skills) with testing, benchmarking, and trigger-tuning capabilities.
OpenClaw's skill-creator already has excellent tooling for creating, packaging, and pub... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T23:14:22Z | 2026-04-23T04:22:52Z | 2026-04-20T04:30:40Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | Binlogo | 7,845,507 | MDQ6VXNlcjc4NDU1MDc= | User | false |
openclaw/openclaw | 4,024,644,171 | I_kwDOQb6kR87v4zJL | 35,039 | https://github.com/openclaw/openclaw/issues/35039 | https://api.github.com/repos/openclaw/openclaw/issues/35039 | Unable to spawn subagents from telegram sessions. | ### Bug type
Regression (worked before, now fails)
### Summary
When gateway.auth.mode is set to "token", agent tool calls (e.g., sessions_spawn, cron, sessions_list) fail with a 10-second timeout and "device-required" error. The web UI (Control UI) works fine because it uses a different authentication path that resp... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T23:06:14Z | 2026-04-23T04:22:53Z | 2026-04-20T04:30:38Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | iceman3k | 16,768,871 | MDQ6VXNlcjE2NzY4ODcx | User | false |
openclaw/openclaw | 4,024,564,457 | I_kwDOQb6kR87v4frp | 35,035 | https://github.com/openclaw/openclaw/issues/35035 | https://api.github.com/repos/openclaw/openclaw/issues/35035 | ClawControl app: 'origin not allowed' for app://localhost even when in allowedOrigins | ## Description
ClawControl desktop app (v1.5.0, Electron) connecting to gateway over Tailscale Serve cannot connect despite `app://localhost` being explicitly set in `gateway.controlUi.allowedOrigins`.
## Environment
- **OpenClaw version:** 2026.3.2
- **ClawControl version:** 1.5.0 (Electron/28.3.3)
- **OS (gateway)... | closed | not_planned | true | 10 | [
"stale"
] | [] | 2026-03-04T22:45:59Z | 2026-04-23T04:22:53Z | 2026-04-20T04:30:36Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | schergr | 59,206 | MDQ6VXNlcjU5MjA2 | User | false |
openclaw/openclaw | 4,024,531,048 | I_kwDOQb6kR87v4Xho | 35,031 | https://github.com/openclaw/openclaw/issues/35031 | https://api.github.com/repos/openclaw/openclaw/issues/35031 | System events (reactions, pins) ignore agent bindings — session key mismatch | ## Summary
System events (reaction_added, reaction_removed, pin_added, pin_removed, member_joined_channel, etc.) are always enqueued with `agent:main:...` session key, ignoring configured `bindings`. This means bound agents never receive these events during heartbeat drain.
## Root Cause
`resolveSlackSystemEventSess... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T22:37:00Z | 2026-04-23T04:22:54Z | 2026-04-20T04:30:35Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | dvpadmin | 19,187,020 | MDQ6VXNlcjE5MTg3MDIw | User | false |
openclaw/openclaw | 4,024,281,444 | I_kwDOQb6kR87v3alk | 35,005 | https://github.com/openclaw/openclaw/issues/35005 | https://api.github.com/repos/openclaw/openclaw/issues/35005 | Mac App 2026.3.2 crashes at startup: SIGABRT in TalkOverlayController.present() | ## Bug: Mac App crashes at startup with SIGABRT in TalkOverlayController
**Version:** Mac App 2026.3.2 (CFBundleVersion 2026030290)
**OS:** macOS 26.2.0 (Darwin 25.2.0)
**Architecture:** ARM64
### Crash Details
Signal: `SIGABRT` (Abort trap: 6)
The app crashes immediately at startup. The crash is triggered by a Swi... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T21:34:09Z | 2026-04-23T04:22:55Z | 2026-04-20T04:30:33Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | ostheimer | 410,420 | MDQ6VXNlcjQxMDQyMA== | User | false |
openclaw/openclaw | 4,024,214,742 | I_kwDOQb6kR87v3KTW | 34,993 | https://github.com/openclaw/openclaw/issues/34993 | https://api.github.com/repos/openclaw/openclaw/issues/34993 | Sub-agent announce routes to wrong channel in multi-channel shared session | ## Bug
When a sub-agent is spawned via `sessions_spawn` from a **webchat** session, the completion announce message is delivered to **Telegram** instead of back to the webchat session that spawned it.
Both channels share the same session (`agent:main:main`). The announce appears to resolve delivery target from the se... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T21:18:39Z | 2026-04-23T04:22:56Z | 2026-04-20T04:30:31Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | drxm1 | 31,032,638 | MDQ6VXNlcjMxMDMyNjM4 | User | false |
openclaw/openclaw | 4,024,206,962 | I_kwDOQb6kR87v3IZy | 34,992 | https://github.com/openclaw/openclaw/issues/34992 | https://api.github.com/repos/openclaw/openclaw/issues/34992 | [BUG] Memory search fails with "missing node:sqlite" despite better-sqlite3 installed (Node.js 22) | ## Description
OpenClaw version: 2026.3.2 (85377a2)
Node.js version: v22.12.0
npm version: 10.9.0
Operating system: Windows 10/11 (10.0.26300)
Installation method: npm global install
## Problem
When trying to use the memory search functionality, OpenClaw reports:
SQLite support is unavailable in this Node runtime (mi... | closed | not_planned | true | 7 | [
"stale"
] | [] | 2026-03-04T21:17:10Z | 2026-04-23T04:22:56Z | 2026-04-20T04:30:29Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | heronerar | 128,018,749 | U_kgDOB6FpPQ | User | false |
openclaw/openclaw | 4,024,200,191 | I_kwDOQb6kR87v3Gv_ | 34,990 | https://github.com/openclaw/openclaw/issues/34990 | https://api.github.com/repos/openclaw/openclaw/issues/34990 | [Bug]:Incident Report: Recovery of OpenClaw Environment v2026.3.2 (7-Hour Service Outage) | ### Bug type
Regression (worked before, now fails)
### Summary
Subject: Incident Report: Recovery of OpenClaw Environment v2026.3.2 (7-Hour Service Outage)
Total Downtime: ~7 Hours Core Issue: Cascading failures triggered by a forced version update, API model deprecation, and mismatched local security signatures. S... | closed | not_planned | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T21:15:45Z | 2026-04-23T04:22:57Z | 2026-04-20T04:30:27Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | SoulMatchedAI | 263,216,835 | U_kgDOD7Beww | User | false |
openclaw/openclaw | 4,024,106,459 | I_kwDOQb6kR87v2v3b | 34,979 | https://github.com/openclaw/openclaw/issues/34979 | https://api.github.com/repos/openclaw/openclaw/issues/34979 | [Bug]: Malformed tool result content block crashes context truncation in a session-permanent loop | ### Bug type
Crash (process/app exits or hangs)
### Summary
A plugin tool handler that returns `undefined`/`void` produces a malformed content block `{type: "text"}` (no `text` property), which crashes the context truncation pipeline on every subsequent message and locks the session into an unrecoverable crash loop.... | closed | completed | true | 5 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-04T20:55:40Z | 2026-04-25T06:36:42Z | 2026-04-20T04:30:26Z | CONTRIBUTOR | null | 20260425T090015Z | 2026-04-25T09:00:15Z | coffeexcoin | 137,969,418 | U_kgDOCDk_Cg | User | false |
openclaw/openclaw | 4,024,048,127 | I_kwDOQb6kR87v2hn_ | 34,971 | https://github.com/openclaw/openclaw/issues/34971 | https://api.github.com/repos/openclaw/openclaw/issues/34971 | [Bug]: Bug report: Two regressions after 2026.3.2 update | ### Bug type
Regression (worked before, now fails)
### Summary
Issue 1: UI webchat replies bleeding into Telegram
Before the update, UI and Telegram were isolated — messages sent from the Control UI stayed in the UI only. After 2026.3.2, any conversation in the UI also sends replies to Telegram, even when Telegram h... | closed | completed | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T20:42:58Z | 2026-04-22T04:31:01Z | 2026-04-19T12:19:17Z | NONE | null | 20260422T060039Z | 2026-04-22T06:00:39Z | ceo94HEHE | 265,638,402 | U_kgDOD9VSAg | User | false |
openclaw/openclaw | 4,023,967,151 | I_kwDOQb6kR87v2N2v | 34,964 | https://github.com/openclaw/openclaw/issues/34964 | https://api.github.com/repos/openclaw/openclaw/issues/34964 | Heartbeat poll messages should not be visible in TUI output | ## Problem
Heartbeat poll messages (the inbound prompt with `<relevant-memories>`, `Read HEARTBEAT.md...`, etc.) are rendered in the TUI chat view. This is confusing — users see a wall of system text that looks like a broken message.
## Expected Behavior
Heartbeat polls and their `HEARTBEAT_OK` responses should be f... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T20:24:11Z | 2026-04-23T04:22:59Z | 2026-04-20T04:30:24Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | nishantkabra77 | 194,645,679 | U_kgDOC5oOrw | User | false |
openclaw/openclaw | 4,023,954,865 | I_kwDOQb6kR87v2K2x | 34,961 | https://github.com/openclaw/openclaw/issues/34961 | https://api.github.com/repos/openclaw/openclaw/issues/34961 | [Bug]: Discord voice replies intermittently fail with no reply despite successful /vc join | ### Bug type
Regression (worked before, now fails)
### Summary
Discord VC join succeeds, but spoken prompts frequently produce no bot voice reply. Behavior is intermittent and can fail repeatedly even after leave/join and restarts.
### Steps to reproduce
1. Configure Discord channel + OpenClaw Discord voice enable... | closed | not_planned | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T20:21:12Z | 2026-04-23T04:23:00Z | 2026-04-20T04:30:21Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | codsworth-dangera | 265,637,932 | U_kgDOD9VQLA | User | false |
openclaw/openclaw | 4,023,906,280 | I_kwDOQb6kR87v1-_o | 34,953 | https://github.com/openclaw/openclaw/issues/34953 | https://api.github.com/repos/openclaw/openclaw/issues/34953 | SPDX 2.3 | https://spdx.org/rdf/ontology/spdx-2-3/ | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T20:10:26Z | 2026-04-23T04:23:00Z | 2026-04-20T04:30:19Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | Pediboi666 | 188,489,565 | U_kgDOCzwfXQ | User | false |
openclaw/openclaw | 4,023,888,491 | I_kwDOQb6kR87v16pr | 34,948 | https://github.com/openclaw/openclaw/issues/34948 | https://api.github.com/repos/openclaw/openclaw/issues/34948 | Feature Request: Add --probe-all-models and --probe-model to openclaw models status | ## Problem
Today, `openclaw models status --probe` appears to run a provider-level probe (typically one representative model per provider).
This is useful for auth/connectivity checks, but it is not enough to confirm that a specific bound model is actually available.
Example impact:
- Agent binds to `provider/model-... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T20:06:14Z | 2026-04-23T04:23:01Z | 2026-04-20T04:30:18Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | xmanrui | 4,094,054 | MDQ6VXNlcjQwOTQwNTQ= | User | false |
openclaw/openclaw | 4,023,851,648 | I_kwDOQb6kR87v1xqA | 34,941 | https://github.com/openclaw/openclaw/issues/34941 | https://api.github.com/repos/openclaw/openclaw/issues/34941 | [BUG] Telegram inbound is dropped in both polling and webhook modes while outbound works (lastInboundAt stays null) | # Incident report — Telegram inbound broken (OpenClaw) — 2026-03-04
## Summary
Telegram outbound works but inbound is never delivered to OpenClaw sessions.
This reproduces across:
- two different bot tokens/bots,
- polling mode and webhook mode,
- multiple clean resets (state wipe, cold restart, single gateway process... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T19:58:02Z | 2026-04-23T04:23:02Z | 2026-04-20T04:30:16Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | obsidian-assistant | 261,743,431 | U_kgDOD5njRw | User | false |
openclaw/openclaw | 4,023,836,565 | I_kwDOQb6kR87v1t-V | 34,938 | https://github.com/openclaw/openclaw/issues/34938 | https://api.github.com/repos/openclaw/openclaw/issues/34938 | feat: lifecycle hooks for agent and subagent events (before/after) | ## Feature Request: Agent Lifecycle Hooks
### Problem
Currently there is no way to automatically execute actions before/after an agent processes a message or spawns a subagent. This forces developers to rely on prompt instructions (unreliable) instead of code.
### Proposed Solution
Add configurable lifecycle hooks ... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T19:54:45Z | 2026-04-23T04:23:03Z | 2026-04-20T04:30:14Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | motor-bot-agent | 259,915,998 | U_kgDOD34A3g | User | false |
openclaw/openclaw | 4,023,810,478 | I_kwDOQb6kR87v1nmu | 34,935 | https://github.com/openclaw/openclaw/issues/34935 | https://api.github.com/repos/openclaw/openclaw/issues/34935 | [Bug]: safeguard compaction makes LLM API call before checking for real messages — 48 unnecessary calls/day on idle sessions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
With compaction.mode: "safeguard", OpenClaw calls the LLM API every ~30 minutes on all active sessions — including isolated cron sessions that contain no real conversation messages. The safeguard then correctly cancels the compaction, but ... | closed | not_planned | true | 5 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-04T19:48:35Z | 2026-04-23T04:23:03Z | 2026-04-20T04:30:12Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | rushter777 | 17,453,883 | MDQ6VXNlcjE3NDUzODgz | User | false |
openclaw/openclaw | 4,023,757,327 | I_kwDOQb6kR87v1aoP | 34,925 | https://github.com/openclaw/openclaw/issues/34925 | https://api.github.com/repos/openclaw/openclaw/issues/34925 | [Bug]: 2026.3.2 CLI crashes on startup — missing discord-api-types/v10.mjs | ## Description
After updating to `2026.3.2`, the OpenClaw CLI fails to start with the following error:
```
[openclaw] Failed to start CLI: Error: ENOENT: no such file or directory, open '/usr/local/lib/node_modules/openclaw/node_modules/discord-api-types/v10.mjs'
at Object.openSync (node:fs:560:18)
at readFil... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T19:35:51Z | 2026-04-23T04:23:04Z | 2026-04-20T04:30:10Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | ted-networkchuck | 221,014,973 | U_kgDODSxrvQ | User | false |
openclaw/openclaw | 4,023,726,425 | I_kwDOQb6kR87v1TFZ | 34,921 | https://github.com/openclaw/openclaw/issues/34921 | https://api.github.com/repos/openclaw/openclaw/issues/34921 | feat: persistent inline buttons / custom keyboard for Telegram channel | ## Feature Request
### Problem
Currently, inline buttons can only be added via the `message` tool with explicit `buttons` parameter. There is no way to:
1. Add a persistent button to every agent response automatically
2. Show a "Stop" button before the agent starts processing
This means users have no quick way to int... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-04T19:28:25Z | 2026-04-23T04:23:05Z | 2026-04-20T04:30:08Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | motor-bot-agent | 259,915,998 | U_kgDOD34A3g | User | false |
openclaw/openclaw | 4,023,706,079 | I_kwDOQb6kR87v1OHf | 34,912 | https://github.com/openclaw/openclaw/issues/34912 | https://api.github.com/repos/openclaw/openclaw/issues/34912 | [Bug] macOS app crashes when using Voice Wake / Push-to-Talk / Talk Mode (VoiceWakeOverlayController.present) | ### Summary
OpenClaw macOS app crashes with SIGABRT when voice overlay flows are triggered.
### Version / Environment
- OpenClaw app: 2026.3.2 (2026030290)
- macOS: 26.4 (25E5207k)
- Hardware: Mac16,8 (Apple Silicon)
### What happens
- App may launch normally.
- Enabling/triggering voice features causes immediate cra... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-04T19:24:02Z | 2026-04-23T04:23:06Z | 2026-04-20T04:30:06Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | Pawel-Dymek | 20,336,805 | MDQ6VXNlcjIwMzM2ODA1 | User | false |
openclaw/openclaw | 4,023,685,565 | I_kwDOQb6kR87v1JG9 | 34,908 | https://github.com/openclaw/openclaw/issues/34908 | https://api.github.com/repos/openclaw/openclaw/issues/34908 | [Bug]: ontrol UI version badge shows 2026.2.24 in 2026.3.2 release | ### Bug type
Regression (worked before, now fails)
### Summary
The Control UI dashboard displays "Version 2026.2.24" after updating to 2026.3.2. The old version string is hardcoded in the pre-built frontend bundle shipped with the npm package.
**Environment:** macOS ARM64 (Lume VM), Node.js 25.6.1, npm global insta... | closed | not_planned | true | 4 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-04T19:19:17Z | 2026-04-23T04:23:07Z | 2026-04-20T04:30:04Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | KOB1Geo | 208,837,020 | U_kgDODHKZnA | User | false |
openclaw/openclaw | 4,023,674,858 | I_kwDOQb6kR87v1Gfq | 34,903 | https://github.com/openclaw/openclaw/issues/34903 | https://api.github.com/repos/openclaw/openclaw/issues/34903 | [Bug]: macOS app crashes with fatal access conflict in TalkOverlayController.present() when voice wake is enabled | ## Description
The macOS desktop app crashes on launch when Voice Wake / Talk Mode is enabled. The crash occurs in `TalkOverlayController.present()` with a `Fatal access conflict detected` error.
## Steps to Reproduce
1. Enable Voice Wake in the OpenClaw Mac app settings
2. Restart the app (or let it restart via Lau... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T19:16:43Z | 2026-04-23T04:23:07Z | 2026-04-20T04:30:03Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | camden-lowrance | 76,054,170 | MDQ6VXNlcjc2MDU0MTcw | User | false |
openclaw/openclaw | 4,023,657,437 | I_kwDOQb6kR87v1CPd | 34,901 | https://github.com/openclaw/openclaw/issues/34901 | https://api.github.com/repos/openclaw/openclaw/issues/34901 | Bug: OPENCLAW_SERVICE_VERSION in LaunchAgent plist not updated on `openclaw update` (macOS) | ## Summary
After running `openclaw update`, the gateway binary is correctly upgraded but the LaunchAgent plist at `~/Library/LaunchAgents/ai.openclaw.gateway.plist` retains the old `OPENCLAW_SERVICE_VERSION` env var. The gateway process falls back to this env var for its reported version, so the Control UI shows the o... | closed | not_planned | true | 6 | [
"stale"
] | [] | 2026-03-04T19:12:44Z | 2026-04-23T04:23:08Z | 2026-04-20T04:30:01Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | tonyjwinter | 2,146,144 | MDQ6VXNlcjIxNDYxNDQ= | User | false |
openclaw/openclaw | 4,023,591,369 | I_kwDOQb6kR87v0yHJ | 34,881 | https://github.com/openclaw/openclaw/issues/34881 | https://api.github.com/repos/openclaw/openclaw/issues/34881 | Feature: /steer as shorthand for /queue steer in main session | ## Summary
When the queue mode is set to `collect` (the default), users currently need to write `/queue steer` inline to inject a message into the active agent run. This is verbose and unintuitive.
## Proposed behavior
Extend `/steer` so that when used **without a sub-agent ID** in the main session, it acts as a sho... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-04T18:58:48Z | 2026-04-23T04:23:09Z | 2026-04-20T04:29:59Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | Christoffer91 | 92,569,289 | U_kgDOBYR-yQ | User | false |
openclaw/openclaw | 4,023,564,365 | I_kwDOQb6kR87v0rhN | 34,875 | https://github.com/openclaw/openclaw/issues/34875 | https://api.github.com/repos/openclaw/openclaw/issues/34875 | [Feature] i18n / customizable subagent notification messages (timed out / finished) | ## Problem
When a subagent completes or times out, OpenClaw pushes a hardcoded English notification to the chat surface (e.g. Telegram):
- `⏱️ Subagent main timed out`
- `✅ Subagent main finished`
These messages are not customizable and cannot be localized. For non-English-speaking users, this creates a jarring expe... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-04T18:52:43Z | 2026-04-23T04:23:10Z | 2026-04-20T04:29:57Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | 0xWinner98 | 164,893,774 | U_kgDOCdQUTg | User | false |
openclaw/openclaw | 4,023,494,638 | I_kwDOQb6kR87v0afu | 34,857 | https://github.com/openclaw/openclaw/issues/34857 | https://api.github.com/repos/openclaw/openclaw/issues/34857 | bug: agents model selection <select> shows wrong primary model on initial load | ## Bug
When opening the **Agents → Overview** tab in the Control UI, the **Model Selection** `<select>` for primary model shows the **wrong model** on initial page load — it shows the first model in the configured list instead of the actual configured primary.
The **Overview read-only row** ("Primary Model") correctl... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-04T18:37:03Z | 2026-04-23T04:23:11Z | 2026-04-20T04:29:55Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | aashari | 4,284,487 | MDQ6VXNlcjQyODQ0ODc= | User | false |
openclaw/openclaw | 4,023,448,493 | I_kwDOQb6kR87v0POt | 34,852 | https://github.com/openclaw/openclaw/issues/34852 | https://api.github.com/repos/openclaw/openclaw/issues/34852 | fix(gateway): node misses events from spawned child agent sessions | ## Bug
When an agent spawns a child agent via `sessions_spawn`, the gateway's node subscription system does not route the child's events to the parent's subscribed nodes.
### Reproduction
1. A node subscribes to `agent:agentlife:main` via `chat.subscribe`
2. `agentlife` spawns a `health` agent via `sessions_spawn`
3... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-04T18:26:25Z | 2026-04-23T04:23:11Z | 2026-04-20T04:29:54Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | carbaj03 | 5,271,767 | MDQ6VXNlcjUyNzE3Njc= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.