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 | 4,377,499,465 | I_kwDOQb6kR88AAAABBOtXSQ | 77,369 | https://github.com/openclaw/openclaw/issues/77369 | https://api.github.com/repos/openclaw/openclaw/issues/77369 | Bug: Telegram "Sifting" progress draft label persists after bot restart | ## Bug Report
**Bug type:** bug
**Summary:** Telegram "Sifting..." progress draft label persists after bot restart
---
**Steps to reproduce:**
1. Configure a Telegram bot with `streaming.mode: "progress"` and `streaming.labels` using the default pool (including "Sifting")
2. Send a message that triggers the agent t... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T14:50:59Z | 2026-05-04T15:29:34Z | 2026-05-04T15:19:49Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Hansen1018 | 61,605,071 | MDQ6VXNlcjYxNjA1MDcx | User | false |
openclaw/openclaw | 4,376,767,881 | I_kwDOQb6kR88AAAABBOAtiQ | 77,324 | https://github.com/openclaw/openclaw/issues/77324 | https://api.github.com/repos/openclaw/openclaw/issues/77324 | Streamed assistant messages not persisted in transcript (Control UI / webchat) | null | closed | not_planned | false | 1 | [] | [] | 2026-05-04T13:00:11Z | 2026-05-04T15:27:04Z | 2026-05-04T15:24:43Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | bogdanverra-crypto | 255,924,763 | U_kgDOD0EaGw | User | false |
openclaw/openclaw | 4,377,599,349 | I_kwDOQb6kR88AAAABBOzddQ | 77,378 | https://github.com/openclaw/openclaw/issues/77378 | https://api.github.com/repos/openclaw/openclaw/issues/77378 | [Bug]: Session rotation creates duplicate sessions with broken delivery context | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When the Mattermost channel is reloaded (e.g., after config change, gateway restart, or channel restart), the gateway creates a **new session** for the same group channel while the **old session remains active*... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-05-04T15:06:30Z | 2026-05-04T22:15:56Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | wurdzy | 281,569,148 | U_kgDOEMhnfA | User | false |
openclaw/openclaw | 4,377,589,754 | I_kwDOQb6kR88AAAABBOy3-g | 77,377 | https://github.com/openclaw/openclaw/issues/77377 | https://api.github.com/repos/openclaw/openclaw/issues/77377 | Bug: Feishu messages concatenate cumulatively after upgrading to 2026.5.2 | ## Bug Report: Feishu messages concatenate cumulatively after upgrading to 2026.5.2
### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
After upgrading to 2026.5.2, Feishu channel messages are being concatenated/cumulative when sent to the OpenClaw agent. Instead of each message being sent in... | closed | completed | false | 1 | [] | [] | 2026-05-04T15:05:03Z | 2026-05-04T22:12:00Z | 2026-05-04T22:12:00Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | EdwinXu123 | 201,543,183 | U_kgDODANODw | User | false |
openclaw/openclaw | 4,377,648,485 | I_kwDOQb6kR88AAAABBO2dZQ | 77,386 | https://github.com/openclaw/openclaw/issues/77386 | https://api.github.com/repos/openclaw/openclaw/issues/77386 | image_generate via OpenAI Codex OAuth hits hardcoded 30s dynamic tool watchdog and loses result | ## Summary
`image_generate` with `openai/gpt-image-2` through the OpenAI Codex OAuth transport starts correctly with a configured 180s timeout, but the Codex dynamic tool wrapper aborts waiting after 30s:
```text
[agent/embedded] codex dynamic tool timeout: tool=image_generate toolTimeoutMs=30000; per-tool-call watch... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T15:14:38Z | 2026-05-04T15:22:17Z | 2026-05-04T15:22:17Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | nightsnowfox | 61,942,995 | MDQ6VXNlcjYxOTQyOTk1 | User | false |
openclaw/openclaw | 4,377,646,460 | I_kwDOQb6kR88AAAABBO2VfA | 77,384 | https://github.com/openclaw/openclaw/issues/77384 | https://api.github.com/repos/openclaw/openclaw/issues/77384 | Regression: Custom commands in routing layer are no longer intercepted/executed before LLM turn in v2026.5.x | ### Description
In version **2026.4.23**, custom commands defined in the routing layer (specifically via plugins like `master-control`) worked perfectly. When a user sent a registered command (e.g., `/bf`, `/asset`, `/nodeseek`), the system would intercept it at the routing level and execute the local handler immediate... | closed | completed | false | 1 | [] | [] | 2026-05-04T15:14:18Z | 2026-05-04T15:23:14Z | 2026-05-04T15:23:07Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | githubdulong | 59,136,159 | MDQ6VXNlcjU5MTM2MTU5 | User | false |
openclaw/openclaw | 4,377,662,189 | I_kwDOQb6kR88AAAABBO3S7Q | 77,387 | https://github.com/openclaw/openclaw/issues/77387 | https://api.github.com/repos/openclaw/openclaw/issues/77387 | [Bug] Feishu: message content concatenates all previous messages instead of processing current message only | ## Bug Report: Feishu message content concatenates previous messages instead of processing only current message
### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
After upgrading to OpenClaw 2026.5.2, Feishu channel exhibits intermittent message concatenation. When a user sends a new message... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T15:16:51Z | 2026-05-04T15:24:19Z | 2026-05-04T15:24:19Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | EdwinXu123 | 201,543,183 | U_kgDODANODw | User | false |
openclaw/openclaw | 4,377,643,768 | I_kwDOQb6kR88AAAABBO2K-A | 77,383 | https://github.com/openclaw/openclaw/issues/77383 | https://api.github.com/repos/openclaw/openclaw/issues/77383 | [Bug]: Feishu @all 消息无法被 respondToMentionAll 正确识别，bot 不回复群聊 @所有人 | ## 问题描述
在飞书群聊中，即使配置了 `respondToMentionAll: true`，bot 仍然不会回复 `@所有人` 的消息。
## 环境
- **OpenClaw 版本**: 2026.5.2 (8b2a6e5)
- **飞书插件版本**: @larksuite/openclaw-lark 2026.4.10
- **操作系统**: macOS
- **安装方式**: npm global
## 配置
```json
{
"channels": {
"feishu": {
"requireMention": true,
"respondToMentionAll": tr... | open | null | false | 1 | [] | [] | 2026-05-04T15:13:49Z | 2026-05-04T15:25:54Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | yiyouguisu | 11,996,269 | MDQ6VXNlcjExOTk2MjY5 | User | false |
openclaw/openclaw | 4,377,716,727 | I_kwDOQb6kR88AAAABBO6n9w | 77,391 | https://github.com/openclaw/openclaw/issues/77391 | https://api.github.com/repos/openclaw/openclaw/issues/77391 | diagnostics.otel.captureContent.* is non-functional — runtime broadcast emits sanitized paramsSummary only | ## Summary
`diagnostics.otel.captureContent.{inputMessages,outputMessages,toolInputs,toolOutputs}` is documented but does nothing. The runtime emits only sanitized `paramsSummary` on the broadcast stream that bundled `diagnostics-otel` consumes — raw `inputMessages`, `toolInput`, and `toolResult` content never appear ... | open | null | false | 2 | [] | [] | 2026-05-04T15:24:39Z | 2026-05-05T01:48:14Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | mjunaidca | 28,400,845 | MDQ6VXNlcjI4NDAwODQ1 | User | false |
openclaw/openclaw | 4,377,715,791 | I_kwDOQb6kR88AAAABBO6kTw | 77,390 | https://github.com/openclaw/openclaw/issues/77390 | https://api.github.com/repos/openclaw/openclaw/issues/77390 | diagnostics-otel: bundled-only origin gate prevents external installs from receiving internal diagnostics | ## Summary
Externally-installed `diagnostics-otel` plugins are silently denied access to internal diagnostics events because the `grantsInternalDiagnostics` capability is gated on `origin === "bundled"`. Plugins installed via `clawhub:` or `npm:` get `origin: "global"` and the gate fails closed — the plugin loads but ... | closed | completed | false | 1 | [] | [] | 2026-05-04T15:24:31Z | 2026-05-04T22:12:07Z | 2026-05-04T22:12:07Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mjunaidca | 28,400,845 | MDQ6VXNlcjI4NDAwODQ1 | User | false |
openclaw/openclaw | 4,377,744,985 | I_kwDOQb6kR88AAAABBO8WWQ | 77,392 | https://github.com/openclaw/openclaw/issues/77392 | https://api.github.com/repos/openclaw/openclaw/issues/77392 | Discord replies have ~10–11s end-to-end latency even for trivial "OK" responses | # Bug Report: Discord replies have ~10–11s end-to-end latency even for trivial `OK` responses
## Summary
OpenClaw Discord bot replies are consistently slow in a minimal test case.
Even when the user asks the agent to only reply `OK`, the end-to-end latency from the Discord user message timestamp to the Discord bot r... | open | null | false | 1 | [] | [] | 2026-05-04T15:28:54Z | 2026-05-04T15:37:11Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | leoge007 | 116,452,300 | U_kgDOBvDrzA | User | false |
openclaw/openclaw | 4,377,923,292 | I_kwDOQb6kR88AAAABBPHO3A | 77,396 | https://github.com/openclaw/openclaw/issues/77396 | https://api.github.com/repos/openclaw/openclaw/issues/77396 | active-memory: Bundled plugin dirName must be a single directory (QQ c2c) | ## Bug 描述active-memory 插件在处理 QQ 私信（c2c）消息时崩溃。**复现步骤：**1. 启用 active-memory 插件2. 通过 QQ 机器人发送私信给飞牛3. active-memory 子代理启动时立即报错**错误日志：**Error: Bundled plugin dirName must be a single directory: c2c:10D4F7C20EF54550C2845B47A3C95803lane task error: lane=main durationMs=4741active-memory: agent=main session=agent:main:main fai... | closed | completed | false | 1 | [] | [] | 2026-05-04T15:59:39Z | 2026-05-04T16:08:01Z | 2026-05-04T16:08:01Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | zhao932 | 267,828,416 | U_kgDOD_a8wA | User | false |
openclaw/openclaw | 4,377,953,999 | I_kwDOQb6kR88AAAABBPJGzw | 77,397 | https://github.com/openclaw/openclaw/issues/77397 | https://api.github.com/repos/openclaw/openclaw/issues/77397 | Image tool fails with Bedrock + `aws-sdk` auth mode — missing carve-out in `requireApiKey` | ## Summary
The `image` tool refuses to run on any Bedrock model when the provider is configured with `auth.mode: "aws-sdk"` and credentials live in `~/.aws/credentials` / env vars rather than being pasted into the config as `apiKey`.
The **text / main-loop** code path already has the correct carve-out — it passes `un... | open | null | false | 1 | [] | [] | 2026-05-04T16:06:30Z | 2026-05-04T16:13:21Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | bstanbury | 159,021,261 | U_kgDOCXp4zQ | User | false |
openclaw/openclaw | 4,377,964,492 | I_kwDOQb6kR88AAAABBPJvzA | 77,398 | https://github.com/openclaw/openclaw/issues/77398 | https://api.github.com/repos/openclaw/openclaw/issues/77398 | Discord channel inbound broken on IPv4-only networks (2026.5.x): undici "happy eyeballs" never falls back from failing AAAA lookup | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from 2026.4.27 to 2026.5.x, Discord channel inbound stops working on networks where discord.com only has A records (no AAAA). The bot can still send messages via REST, but the gateway WebSocket never receives... | closed | completed | false | 2 | [
"bug",
"regression"
] | [] | 2026-05-04T16:08:50Z | 2026-05-05T02:32:01Z | 2026-05-05T02:32:01Z | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | mwinschel | 18,651,387 | MDQ6VXNlcjE4NjUxMzg3 | User | false |
openclaw/openclaw | 4,377,984,968 | I_kwDOQb6kR88AAAABBPK_yA | 77,400 | https://github.com/openclaw/openclaw/issues/77400 | https://api.github.com/repos/openclaw/openclaw/issues/77400 | doctor --fix should not delete auth.profiles entries for providers with configured API keys | ## Summary
`openclaw doctor --fix` silently deletes entries from `auth.profiles` — including `openai:default` — even when the corresponding provider's API key is present and valid in `models.providers` (or the equivalent config location). This causes a complete fallback-chain failure with no error surfaced to the user... | closed | completed | false | 3 | [] | [] | 2026-05-04T16:12:55Z | 2026-05-04T22:30:37Z | 2026-05-04T22:30:37Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | axonrelaybot | 275,827,808 | U_kgDOEHDMYA | User | false |
openclaw/openclaw | 4,377,986,475 | I_kwDOQb6kR88AAAABBPLFqw | 77,401 | https://github.com/openclaw/openclaw/issues/77401 | https://api.github.com/repos/openclaw/openclaw/issues/77401 | [Bug]: Telegram: final reply delivered before last streaming progress messages on tool-heavy turns (race condition) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When an agent performs multiple tool calls with `streaming.mode: "partial"`, the final
assembled reply occasionally arrives in Telegram **before** the last few streaming progress
messages (Exec, Process, etc.... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-05-04T16:13:14Z | 2026-05-04T16:22:53Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | SheraSnow | 281,348,569 | U_kgDOEMUJ2Q | User | false |
openclaw/openclaw | 4,377,708,500 | I_kwDOQb6kR88AAAABBO6H1A | 77,389 | https://github.com/openclaw/openclaw/issues/77389 | https://api.github.com/repos/openclaw/openclaw/issues/77389 | Bug: Telegram progress draft "Sifting" persists and triggers at every step after restart | **Bug type:** bug
**Summary:** Telegram progress draft label (e.g., "Sifting...") persists after bot restart and triggers at every execution step instead of being replaced by final response.
---
## Steps to reproduce
1. Configure Telegram bot with streaming.mode: "partial" (default mode)
2. Send any command that tr... | open | null | false | 1 | [] | [] | 2026-05-04T15:23:32Z | 2026-05-04T16:26:12Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Hansen1018 | 61,605,071 | MDQ6VXNlcjYxNjA1MDcx | User | false |
openclaw/openclaw | 4,378,027,140 | I_kwDOQb6kR88AAAABBPNkhA | 77,404 | https://github.com/openclaw/openclaw/issues/77404 | https://api.github.com/repos/openclaw/openclaw/issues/77404 | [Bug]: | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
The bug #72547 is not fixed yet.
With HTTP_PROXY env variable set, after scan QR code, from openclaw dashboard and openclaw status output, It's said whatsapp **linked** but never **connected**.
- @openclaw/whatsapp "versi... | closed | not_planned | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T16:21:58Z | 2026-05-04T20:04:36Z | 2026-05-04T20:04:36Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | mebusw | 760,587 | MDQ6VXNlcjc2MDU4Nw== | User | false |
openclaw/openclaw | 4,378,093,295 | I_kwDOQb6kR88AAAABBPRm7w | 77,407 | https://github.com/openclaw/openclaw/issues/77407 | https://api.github.com/repos/openclaw/openclaw/issues/77407 | Bug: image_generate fails with EPERM fsync on Windows | ## Description
`image_generate` consistently fails with `EPERM: operation not permitted, fsync` when trying to save generated images to disk on Windows.
## Environment
- OS: Windows 10.0.26200 (x64)
- Node: v24.14.0
- OpenClaw: 2026.5.3-1 (stable)
- Channel: Telegram (direct)
## Steps to Reproduce
1. Call `image_gen... | closed | completed | false | 0 | [] | [] | 2026-05-04T16:36:10Z | 2026-05-04T16:37:09Z | 2026-05-04T16:37:09Z | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | poiskgit | 129,569,601 | U_kgDOB7kTQQ | User | false |
openclaw/openclaw | 4,378,065,100 | I_kwDOQb6kR88AAAABBPP4zA | 77,405 | https://github.com/openclaw/openclaw/issues/77405 | https://api.github.com/repos/openclaw/openclaw/issues/77405 | Control UI Webchat: Assistant responses render as fragmented bubbles during tool-calling turns | # Issue: Control UI Webchat 响应渲染碎片化
## 标题
Control UI Webchat: Assistant responses render as fragmented bubbles during tool-calling turns
## 描述
### 问题
当 assistant 在一次回复中交替输出文字和调用工具时，Control UI Webchat 会把每个文字片段渲染成独立的气泡，导致用户看到大量碎片化的消息块堆叠在一起。
### 复现步骤
1. 打开 Control UI Webchat
2. 发送一个需要多步工具调用的请求（例如：审查本地文件并修复多个问题）
3. 观察 ... | closed | not_planned | false | 3 | [] | [] | 2026-05-04T16:30:05Z | 2026-05-05T06:23:56Z | 2026-05-05T06:22:07Z | NONE | null | 20260505T090045Z | 2026-05-05T09:00:45Z | VeryPond | 182,646,517 | U_kgDOCuL29Q | User | false |
openclaw/openclaw | 4,378,096,337 | I_kwDOQb6kR88AAAABBPRy0Q | 77,408 | https://github.com/openclaw/openclaw/issues/77408 | https://api.github.com/repos/openclaw/openclaw/issues/77408 | Webchat composer + mobile controls clipped in iOS 26.x Safari standalone PWA mode (env(safe-area-inset-bottom) under-reports) | ## Summary
In Safari "Add to Home Screen" standalone (PWA) mode on iOS 26.x, the control-ui webchat composer and its mobile controls row (`.chat-mobile-controls-wrapper`) are clipped below the iPhone home indicator / safe area. The action buttons (model select, thinking toggle, attach, etc.) below the text input are h... | open | null | false | 1 | [] | [] | 2026-05-04T16:36:50Z | 2026-05-04T16:45:00Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | bstanbury | 159,021,261 | U_kgDOCXp4zQ | User | false |
openclaw/openclaw | 4,378,196,373 | I_kwDOQb6kR88AAAABBPX5lQ | 77,419 | https://github.com/openclaw/openclaw/issues/77419 | https://api.github.com/repos/openclaw/openclaw/issues/77419 | Moved to private security disclosure | This report has been withdrawn from the public issue tracker and moved to the private security disclosure flow per the repository security policy.
No secrets were included in the original report, but the topic may be security-sensitive because it concerns credential-containing config file permissions. Please triage th... | closed | not_planned | false | 0 | [] | [] | 2026-05-04T16:55:28Z | 2026-05-04T16:57:59Z | 2026-05-04T16:57:59Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Kaze310 | 183,121,490 | U_kgDOCuo2Ug | User | false |
openclaw/openclaw | 4,378,118,189 | I_kwDOQb6kR88AAAABBPTILQ | 77,409 | https://github.com/openclaw/openclaw/issues/77409 | https://api.github.com/repos/openclaw/openclaw/issues/77409 | stripManagedDailyDreamingLines aborts inside its own REM block — boundary regex matches `### Reflections` heading | ## Summary
`isManagedDailyDreamingBoundary` in `extensions/memory-core/src/dreaming-phases.ts` uses the regex `/^#{1,6}\s+/` to detect markdown-heading boundaries when stripping a managed daily dreaming block. That regex is too broad — it matches the REM block's *own* internal `### Reflections` (and similar) heading, ... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T16:41:05Z | 2026-05-04T16:58:54Z | 2026-05-04T16:58:54Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | eric-basketbot | 266,982,593 | U_kgDOD-nUwQ | User | false |
openclaw/openclaw | 4,378,176,070 | I_kwDOQb6kR88AAAABBPWqRg | 77,414 | https://github.com/openclaw/openclaw/issues/77414 | https://api.github.com/repos/openclaw/openclaw/issues/77414 | [Bug]: Gateway PID + children cannot reach LAN hosts on macOS when exposed via Tailscale Funnel (per-PID NE egress filter) | ## Summary
On macOS, when the OpenClaw gateway is exposed via **Tailscale Funnel** (`tailscale.mode = "funnel"` / `tailscale serve --funnel` over `*.ts.net`), the gateway process and **every child it spawns** silently lose the ability to reach hosts on the local LAN (e.g. `192.168.1.0/24`), returning `No route to host... | open | null | false | 1 | [] | [] | 2026-05-04T16:52:03Z | 2026-05-04T16:58:55Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | rafaelreis-r | 57,492,577 | MDQ6VXNlcjU3NDkyNTc3 | User | false |
openclaw/openclaw | 4,378,202,903 | I_kwDOQb6kR88AAAABBPYTFw | 77,422 | https://github.com/openclaw/openclaw/issues/77422 | https://api.github.com/repos/openclaw/openclaw/issues/77422 | Subagent announcement output not pushed to webchat UI when parent orchestrator is idle (queue-drained run) | ## Summary
When a subagent completes and announces back to a parent webchat orchestrator session that's currently idle, the orchestrator processes the announcement and writes its response to `chat.history` on disk — but the webchat UI never receives a websocket push. The user sees the new message only after a UI recon... | closed | completed | false | 1 | [] | [] | 2026-05-04T16:56:44Z | 2026-05-04T22:31:08Z | 2026-05-04T22:31:08Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | oiGaDio | 265,029,916 | U_kgDOD8wJHA | User | false |
openclaw/openclaw | 4,378,224,775 | I_kwDOQb6kR88AAAABBPZohw | 77,424 | https://github.com/openclaw/openclaw/issues/77424 | https://api.github.com/repos/openclaw/openclaw/issues/77424 | [Bug] Feishu plugin JS files missing from Docker image v2026.5.3 | ## Description
The Feishu (飞书/Lark) channel plugin is missing its JavaScript files in the official Docker image `ghcr.io/openclaw/openclaw:2026.5.3-amd64`.
## Environment
- **OpenClaw Version:** 2026.5.3
- **Docker Image:** `ghcr.io/openclaw/openclaw:2026.5.3-amd64`
- **Platform:** linux/amd64 (Debian 12)
## Proble... | closed | completed | false | 2 | [] | [
"vincentkoc"
] | 2026-05-04T17:00:42Z | 2026-05-04T22:11:16Z | 2026-05-04T22:11:16Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | roamer-remote | 267,348,107 | U_kgDOD-9oiw | User | false |
openclaw/openclaw | 4,378,254,359 | I_kwDOQb6kR88AAAABBPbcFw | 77,428 | https://github.com/openclaw/openclaw/issues/77428 | https://api.github.com/repos/openclaw/openclaw/issues/77428 | [Bug]: macOS LaunchAgent repair kickstarts already-loaded running gateway | ## Summary
On macOS, `repairLaunchAgentBootstrap()` still appears to kickstart the OpenClaw gateway LaunchAgent when `launchctl bootstrap` reports that the service is already loaded. If the service is already loaded and running, repair should return success instead of forcing a kickstart/restart.
This is adjacent to,... | open | null | false | 1 | [] | [] | 2026-05-04T17:05:40Z | 2026-05-04T17:12:39Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ramitrkar-hash | 280,062,852 | U_kgDOELFrhA | User | false |
openclaw/openclaw | 4,378,254,842 | I_kwDOQb6kR88AAAABBPbd-g | 77,429 | https://github.com/openclaw/openclaw/issues/77429 | https://api.github.com/repos/openclaw/openclaw/issues/77429 | [Bug]: Discord multi-account startup should prioritize default/main account | ## Summary
With multiple Discord accounts configured, OpenClaw starts accounts in config/list order. In a multi-account setup where `main`/Jarvis is the primary visible bot and specialist accounts such as Billy/Farber/Lumbergh/Maverick/Scout are secondary, the primary account can start after secondary accounts.
On `2... | open | null | false | 1 | [] | [] | 2026-05-04T17:05:44Z | 2026-05-04T17:13:59Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | ramitrkar-hash | 280,062,852 | U_kgDOELFrhA | User | false |
openclaw/openclaw | 4,378,250,532 | I_kwDOQb6kR88AAAABBPbNJA | 77,426 | https://github.com/openclaw/openclaw/issues/77426 | https://api.github.com/repos/openclaw/openclaw/issues/77426 | [Bug]:sessions_yield: always returns "No session context" on MCP/claude-cli agent runtime path (gateway tool resolver missing sessionId + onYield) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
sessions_yield always returns "No session context" on MCP/claude-cli agent runtime path — gateway tool resolver missing sessionId and onYield
### Steps to reproduce
1. Configure agents.defaults.agentRuntime.id = "claude-cl... | open | null | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T17:04:59Z | 2026-05-04T17:14:24Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | jarvisagimuspicard-hub | 281,705,955 | U_kgDOEMp94w | User | false |
openclaw/openclaw | 4,378,254,973 | I_kwDOQb6kR88AAAABBPbefQ | 77,430 | https://github.com/openclaw/openclaw/issues/77430 | https://api.github.com/repos/openclaw/openclaw/issues/77430 | [Bug]: sessions_spawn can drop delegated task/context when agent has systemPromptOverride | ## Summary
After upgrading to OpenClaw `2026.5.3-1`, `sessions_spawn` can create a subagent task with the correct task persisted in the task DB, but the actual child session may receive only the generic bootstrap message and/or lose the generated subagent role context when the target agent has a configured system prom... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T17:05:46Z | 2026-05-04T22:31:14Z | 2026-05-04T22:31:14Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | ramitrkar-hash | 280,062,852 | U_kgDOELFrhA | User | false |
openclaw/openclaw | 4,378,357,499 | I_kwDOQb6kR88AAAABBPhu-w | 77,433 | https://github.com/openclaw/openclaw/issues/77433 | https://api.github.com/repos/openclaw/openclaw/issues/77433 | Bug: Telegram/native sessions repeat prior turn or duplicate user input after retry/compaction | # Bug: Telegram/native agent sessions can repeat prior turn content or replay current user input after context overflow/compaction
## Summary
Leon reports a recurring cross-agent issue: when he asks a new question or gives a new instruction in Telegram, the assistant often repeats the previous answer/task and only the... | closed | completed | false | 1 | [] | [] | 2026-05-04T17:23:07Z | 2026-05-04T17:32:21Z | 2026-05-04T17:32:21Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | leonlushgit | 256,653,063 | U_kgDOD0w3Bw | User | false |
openclaw/openclaw | 4,378,368,161 | I_kwDOQb6kR88AAAABBPiYoQ | 77,434 | https://github.com/openclaw/openclaw/issues/77434 | https://api.github.com/repos/openclaw/openclaw/issues/77434 | [Bug]: /new command in webchat creates a parallel child session instead of resetting the current session | ## Bug Description
When `openclaw.json` is configured with:
```json
"session": {
"dmScope": "main"
}
```
the `/new` command is expected to **reset the current session** and start fresh within the same session context — meaning the existing session is cleared and becomes a new empty session. This behavior works cor... | open | null | false | 2 | [] | [] | 2026-05-04T17:24:58Z | 2026-05-05T07:16:28Z | null | NONE | null | 20260505T090045Z | 2026-05-05T09:00:45Z | jwong-art | 269,971,191 | U_kgDOEBdu9w | User | false |
openclaw/openclaw | 4,378,353,164 | I_kwDOQb6kR88AAAABBPheDA | 77,431 | https://github.com/openclaw/openclaw/issues/77431 | https://api.github.com/repos/openclaw/openclaw/issues/77431 | Prompt cache hit drops to 0% on chat continuations in 5.3-1 — dynamic content lands in system prompt | ## Summary
After upgrading from `2026.4.23` → `2026.5.3-1`, prompt-cache hit rate on chat-continuation requests dropped from ~98% to **0%**, causing per-request cost to jump ~12× (e.g. $0.10 → $1.20 on Opus). Tool-loop continuations are unaffected. The cause: dynamic content (recall snippets, conversation reply contex... | closed | completed | false | 1 | [] | [] | 2026-05-04T17:22:25Z | 2026-05-04T23:54:18Z | 2026-05-04T23:54:18Z | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Udjin79 | 68,610,242 | MDQ6VXNlcjY4NjEwMjQy | User | false |
openclaw/openclaw | 4,378,368,543 | I_kwDOQb6kR88AAAABBPiaHw | 77,435 | https://github.com/openclaw/openclaw/issues/77435 | https://api.github.com/repos/openclaw/openclaw/issues/77435 | [Feature]: Plugin SDK — fail-fast or document hook routing for non-capability plugins | ### Summary
Plugin hooks like `before_dispatch` registered by non-capability-mode plugins silently no-op on the messaging-dispatch path. The plugin's `register()` function itself appears to never be invoked in the gateway's chat-path runtime. Asking for either fail-fast at registration time or documentation of the act... | open | null | false | 1 | [] | [] | 2026-05-04T17:25:02Z | 2026-05-04T17:33:57Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | cm97383 | 215,208,270 | U_kgDODNPRTg | User | false |
openclaw/openclaw | 4,378,378,867 | I_kwDOQb6kR88AAAABBPjCcw | 77,436 | https://github.com/openclaw/openclaw/issues/77436 | https://api.github.com/repos/openclaw/openclaw/issues/77436 | [Bug]: Docker-setup.sh, does not work after 04.29 version | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Creating from scratch a docker container with ./docker-setup.sh and configuring it correctly with API key...etc for Ollama cloud does not work in any way for version after 04.29. The gateway starts but cannot communicate wi... | closed | completed | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T17:26:43Z | 2026-05-04T23:03:53Z | 2026-05-04T23:03:53Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Disruptor85 | 10,594,050 | MDQ6VXNlcjEwNTk0MDUw | User | false |
openclaw/openclaw | 4,378,184,083 | I_kwDOQb6kR88AAAABBPXJkw | 77,416 | https://github.com/openclaw/openclaw/issues/77416 | https://api.github.com/repos/openclaw/openclaw/issues/77416 | [Bug] @openclaw/discord 2026.5.3-1 channel still skipped silently with env-backed SecretRef token (regression of #76371; fix #76449 missed dist/ sidecar layout) | ## Bug
`@openclaw/discord@2026.5.3` (shipped as `openclaw@2026.5.3-1` on npm) **still fails to start the Discord channel** when `channels.discord.token` is configured as an env-backed SecretRef, despite issue #76371 being closed by merged PR #76449 ("fix(secretrefs): resolve external channel contracts").
Failure mode... | closed | completed | false | 5 | [] | [
"vincentkoc"
] | 2026-05-04T16:53:22Z | 2026-05-04T21:57:30Z | 2026-05-04T21:57:30Z | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mogglemoss | 10,966,195 | MDQ6VXNlcjEwOTY2MTk1 | User | false |
openclaw/openclaw | 4,378,085,972 | I_kwDOQb6kR88AAAABBPRKVA | 77,406 | https://github.com/openclaw/openclaw/issues/77406 | https://api.github.com/repos/openclaw/openclaw/issues/77406 | [Bug]: v2026.5.3 systemd gateway install/doctor repair still embeds OPENCLAW_GATEWAY_TOKEN | # Bug type
Behavior bug (incorrect output/state without crash)
# Beta release blocker
No
# Summary
On Ubuntu with a systemd user gateway service, OpenClaw `2026.5.3 (06d46f7)` still generates a service unit containing inline `Environment=OPENCLAW_GATEWAY_TOKEN=...` after both doctor's recommended gateway service r... | closed | completed | false | 3 | [] | [] | 2026-05-04T16:34:34Z | 2026-05-04T18:37:24Z | 2026-05-04T18:37:24Z | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | brokemac79 | 255,583,030 | U_kgDODzvjNg | User | false |
openclaw/openclaw | 4,378,412,854 | I_kwDOQb6kR88AAAABBPlHNg | 77,439 | https://github.com/openclaw/openclaw/issues/77439 | https://api.github.com/repos/openclaw/openclaw/issues/77439 | [Feature]: Configurable trusted-agent allowlist for gateway config.patch/apply | ### Summary
Add an operator-configurable allowlist for agent-driven gateway config.patch/config.apply writes, so trusted personal-assistant deployments can safely delegate non-sensitive config changes without locally patching OpenClaw.
### Problem to solve
The agent-facing gateway tool intentionally fail-closes conf... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-05-04T17:32:23Z | 2026-05-04T17:40:19Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,378,390,340 | I_kwDOQb6kR88AAAABBPjvRA | 77,438 | https://github.com/openclaw/openclaw/issues/77438 | https://api.github.com/repos/openclaw/openclaw/issues/77438 | before_prompt_build hooks not firing for path-loaded plugins | # Issue Report: before_prompt_build Hooks Not Firing for Path-Loaded Plugins
## Summary
Plugins that register `before_prompt_build` hooks via `api.on()` successfully load and register their hooks, but the hooks are never invoked during prompt building. The root cause appears to be three related bugs in OpenClaw's plu... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T17:28:38Z | 2026-05-04T17:42:40Z | 2026-05-04T17:42:40Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | SweetSophia | 44,297,511 | MDQ6VXNlcjQ0Mjk3NTEx | User | false |
openclaw/openclaw | 4,378,423,107 | I_kwDOQb6kR88AAAABBPlvQw | 77,440 | https://github.com/openclaw/openclaw/issues/77440 | https://api.github.com/repos/openclaw/openclaw/issues/77440 | [Bug]: WebChat model selector shows global default instead of agent-specific default model | ## Bug Description
When an agent has its own `model.primary` configured, the WebChat / dashboard model selector displays the wrong model name for the `default` option.
The backend appears to resolve and run the correct agent-specific default model, but the UI label for `default` is resolved from `agents.defaults.mode... | open | null | false | 2 | [] | [] | 2026-05-04T17:34:02Z | 2026-05-05T04:43:41Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | jwong-art | 269,971,191 | U_kgDOEBdu9w | User | false |
openclaw/openclaw | 4,378,437,550 | I_kwDOQb6kR88AAAABBPmnrg | 77,441 | https://github.com/openclaw/openclaw/issues/77441 | https://api.github.com/repos/openclaw/openclaw/issues/77441 | [Bug]: `plugins.allow` auto-generates wrong plugin IDs for built-in channel aliases | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When gateway restarts, `plugins.allow` automatically adds entries `wecom` and `yuanbao` — but these are **channel IDs**, not plugin IDs. The correct plugin IDs should be `wecom-openclaw-plugin` and `openclaw-pl... | closed | completed | false | 1 | [
"bug"
] | [
"vincentkoc"
] | 2026-05-04T17:36:19Z | 2026-05-04T22:15:20Z | 2026-05-04T22:15:20Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | avatasia | 198,357 | MDQ6VXNlcjE5ODM1Nw== | User | false |
openclaw/openclaw | 4,378,470,246 | I_kwDOQb6kR88AAAABBPonZg | 77,444 | https://github.com/openclaw/openclaw/issues/77444 | https://api.github.com/repos/openclaw/openclaw/issues/77444 | [Bug] FAL image generation provider: FAL_KEY env var not resolved, auth profile rejected | ## Bug Description
The FAL image generation provider (`fal-ai/flux/dev`) is registered and appears in `openclaw plugins list` as enabled, but fails to generate images with error: `No image-generation provider registered for fal-ai`.
Two authentication issues are observed:
1. **FAL_KEY env var not resolved**: The plu... | open | null | false | 1 | [] | [] | 2026-05-04T17:42:07Z | 2026-05-04T17:49:24Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | AS76 | 22,101,267 | MDQ6VXNlcjIyMTAxMjY3 | User | false |
openclaw/openclaw | 4,378,464,715 | I_kwDOQb6kR88AAAABBPoRyw | 77,443 | https://github.com/openclaw/openclaw/issues/77443 | https://api.github.com/repos/openclaw/openclaw/issues/77443 | [Bug]: WhatsApp event loop blocked (eventLoopDelayMaxMs=12088ms) on first inbound message — 2026.5.3-1 Windows | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading to 2026.5.3-1 on Windows 11, the WhatsApp plugin blocks the Node.js event loop (eventLoopDelayMaxMs=12088ms, eventLoopUtilization=0.999) immediately after processing the first inbound message, permanently sta... | open | null | false | 3 | [
"bug",
"regression"
] | [] | 2026-05-04T17:41:08Z | 2026-05-05T09:57:48Z | null | NONE | null | 20260505T120033Z | 2026-05-05T12:00:33Z | zaff70 | 272,942,594 | U_kgDOEETGAg | User | false |
openclaw/openclaw | 4,378,489,530 | I_kwDOQb6kR88AAAABBPpyug | 77,447 | https://github.com/openclaw/openclaw/issues/77447 | https://api.github.com/repos/openclaw/openclaw/issues/77447 | [Feature]: Add a memory hygiene doctor and sanitizer | ## Summary
Add a configurable memory/transcript hygiene doctor and sanitizer for unsafe persisted artifacts such as raw local paths, file URLs, delivery directives, and internal tool-output leakage.
## Problem to solve
Long-running agents can persist text that was useful during execution but should not become durabl... | open | null | false | 1 | [] | [] | 2026-05-04T17:45:41Z | 2026-05-04T17:53:57Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | sercada | 24,389,792 | MDQ6VXNlcjI0Mzg5Nzky | User | false |
openclaw/openclaw | 4,378,489,541 | I_kwDOQb6kR88AAAABBPpyxQ | 77,448 | https://github.com/openclaw/openclaw/issues/77448 | https://api.github.com/repos/openclaw/openclaw/issues/77448 | [Feature]: Add a generic outbound artifact delivery guard | ## Summary
Add a channel-neutral outbound artifact delivery guard that turns validated generated file/media references into real attachments and strips unsafe raw path/directive text from the final chat.
## Problem to solve
File-generating agent workflows can end with text such as `MEDIA:./exports/report.xlsx`, `Fil... | open | null | false | 1 | [] | [] | 2026-05-04T17:45:41Z | 2026-05-04T17:54:35Z | null | CONTRIBUTOR | null | 20260504T180039Z | 2026-05-04T18:00:39Z | sercada | 24,389,792 | MDQ6VXNlcjI0Mzg5Nzky | User | false |
openclaw/openclaw | 4,378,646,756 | I_kwDOQb6kR88AAAABBPzY5A | 77,455 | https://github.com/openclaw/openclaw/issues/77455 | https://api.github.com/repos/openclaw/openclaw/issues/77455 | [Bug]: config.patch response returns in-memory config instead of persisted disk state | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from 2026.4.23 to 2026.4.25, config.patch response returns the in-memory `validated.config` instead of what was actually written to disk by `writeConfigFile()`. Since `writeConfigFile()` transforms config via... | open | null | false | 1 | [
"bug"
] | [] | 2026-05-04T18:13:13Z | 2026-05-04T18:21:19Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | guscamara | 146,099,621 | U_kgDOCLVNpQ | User | false |
openclaw/openclaw | 4,378,715,767 | I_kwDOQb6kR88AAAABBP3mdw | 77,458 | https://github.com/openclaw/openclaw/issues/77458 | https://api.github.com/repos/openclaw/openclaw/issues/77458 | openclaw security audit --deep does not propagate SIGTERM to its worker child; orphan reaches PPID=1 at ~96% CPU | ## Summary
When `openclaw security audit --deep` is killed (via `process kill <session>` or external SIGTERM to the launcher), it does not propagate the signal to its forked `openclaw-security` worker child. The worker reparents to init (PPID=1) and continues at ~96% CPU, hammering the gateway and event loop until man... | closed | completed | false | 1 | [] | [
"kevinlin-openai"
] | 2026-05-04T18:25:31Z | 2026-05-04T22:28:51Z | 2026-05-04T22:28:51Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jaikharbanda | 203,336,087 | U_kgDODB6plw | User | false |
openclaw/openclaw | 4,378,702,074 | I_kwDOQb6kR88AAAABBP2w-g | 77,457 | https://github.com/openclaw/openclaw/issues/77457 | https://api.github.com/repos/openclaw/openclaw/issues/77457 | Discord requireMention broken in multi-account setup - ALL messages skipped as no-mention | **Bug Report: Discord `requireMention` broken in multi-account setup - ALL messages skipped as `no-mention`**
**Version:** OpenClaw 2026.5.3-1 (2eae30e)
**Discord Plugin:** Updated today (May 4, 2026 ~14:09)
**Describe the bug**
After updating the Discord plugin today, ALL guild messages in configured channels are be... | open | null | false | 1 | [] | [] | 2026-05-04T18:22:57Z | 2026-05-04T18:37:33Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | zcode-apps | 87,060,116 | MDQ6VXNlcjg3MDYwMTE2 | User | false |
openclaw/openclaw | 4,378,779,937 | I_kwDOQb6kR88AAAABBP7hIQ | 77,461 | https://github.com/openclaw/openclaw/issues/77461 | https://api.github.com/repos/openclaw/openclaw/issues/77461 | [Bug]: plugin-runtime-deps sandbox missing json5 breaks memory_search | # OpenClaw bug report draft — plugin-runtime-deps missing json5
Date: 2026-05-04
OpenClaw version: v2026.5.3-1
Platform: macOS/Darwin arm64, Node v25.9.0
Gateway: local loopback 127.0.0.1:18789
## Summary
`memory_search` failed with a missing `json5` package error from the generated OpenClaw plugin runtime dependenc... | closed | completed | false | 1 | [] | [
"vincentkoc"
] | 2026-05-04T18:36:23Z | 2026-05-04T23:31:51Z | 2026-05-04T23:31:50Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | PalmetteLegacy-hash | 271,649,527 | U_kgDOEDEK9w | User | false |
openclaw/openclaw | 4,378,806,512 | I_kwDOQb6kR88AAAABBP9I8A | 77,467 | https://github.com/openclaw/openclaw/issues/77467 | https://api.github.com/repos/openclaw/openclaw/issues/77467 | [Bug] MiniMax Portal OAuth token cannot auto-refresh — refreshOAuth method not implemented | ## Bug Description
**MiniMax Portal OAuth tokens cannot be automatically refreshed**, causing "No credentials found" errors once the access token expires (~2 hours after initial login). Users must manually re-authenticate each time.
## Steps to Reproduce
1. Configure MiniMax Portal OAuth via `openclaw auth` CLI
2. W... | open | null | false | 3 | [] | [] | 2026-05-04T18:40:53Z | 2026-05-04T18:48:51Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | rongtianhua | 45,085,618 | MDQ6VXNlcjQ1MDg1NjE4 | User | false |
openclaw/openclaw | 4,378,893,692 | I_kwDOQb6kR88AAAABBQCdfA | 77,475 | https://github.com/openclaw/openclaw/issues/77475 | https://api.github.com/repos/openclaw/openclaw/issues/77475 | [Feature]: Intelligent file share | ### Summary
Agents often create files they believe they’re sharing with you by simply providing the file path on their local machine. If you’re not working directly on that machine, you won’t have access to them unless you request they be emailed. This process could be much smarter—depending on the channel, files coul... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-05-04T18:55:18Z | 2026-05-04T19:04:30Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | najef1979-code | 233,555,716 | U_kgDODevHBA | User | false |
openclaw/openclaw | 4,378,814,684 | I_kwDOQb6kR88AAAABBP9o3A | 77,468 | https://github.com/openclaw/openclaw/issues/77468 | https://api.github.com/repos/openclaw/openclaw/issues/77468 | [Bug]: | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
All AutoGLM skills (websearch, open-link, generate-image, image-edit, search-image, deepresearch) return error 410000 "This feature is available after logging in." despite having an active Top-tier subscription. The JWT toke... | closed | not_planned | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T18:42:12Z | 2026-05-04T19:05:24Z | 2026-05-04T19:05:24Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ierofant-boop | 264,947,139 | U_kgDOD8rFww | User | false |
openclaw/openclaw | 4,378,994,492 | I_kwDOQb6kR88AAAABBQInPA | 77,480 | https://github.com/openclaw/openclaw/issues/77480 | https://api.github.com/repos/openclaw/openclaw/issues/77480 | [Bug]: Subagent completion can be lost for active requester sessions when queueEmbeddedPiMessage(...) fails before direct announce fallback | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
A spawned subagent run can complete successfully in its own session artifacts, but its completion never reaches the active requester session when `expectsCompletionMessage=true`; after `sessions_yield(...)`, th... | closed | completed | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-05-04T19:13:35Z | 2026-05-04T19:22:36Z | 2026-05-04T19:22:30Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | daleforbes | 11,342,431 | MDQ6VXNlcjExMzQyNDMx | User | false |
openclaw/openclaw | 4,378,890,322 | I_kwDOQb6kR88AAAABBQCQUg | 77,474 | https://github.com/openclaw/openclaw/issues/77474 | https://api.github.com/repos/openclaw/openclaw/issues/77474 | [Bug]: tool-loop detection disabled by default — agent looped ~50 times across context overflow + auto-compaction (2026-05-04 incident, 3-PR plan attached) | ## Summary
`tools.loopDetection.enabled` defaults to `false`. The detector module exists with a complete, tested implementation (`generic_repeat`, `known_poll_no_progress`, `ping_pong`, `unknown_tool_repeat`, `global_circuit_breaker`) and conservative thresholds (warn=10, critical=20, circuit-breaker=30 — all in `src/... | closed | not_planned | false | 5 | [] | [] | 2026-05-04T18:54:42Z | 2026-05-05T01:17:34Z | 2026-05-05T01:17:34Z | CONTRIBUTOR | null | 20260505T030055Z | 2026-05-05T03:00:55Z | efpiva | 6,392,374 | MDQ6VXNlcjYzOTIzNzQ= | User | false |
openclaw/openclaw | 4,379,102,667 | I_kwDOQb6kR88AAAABBQPNyw | 77,494 | https://github.com/openclaw/openclaw/issues/77494 | https://api.github.com/repos/openclaw/openclaw/issues/77494 | [Bug] Feishu image download fails on Windows with EPERM: operation not permitted, fsync | ## Bug Description
When receiving images via Feishu on Windows, the Gateway fails to download and save images with the error:
`
feishu: failed to download image media: Error: EPERM: operation not permitted, fsync
`
## Environment
- **Platform**: Windows 10/11 (NTFS)
- **Node.js**: v22.22.2
- **OpenClaw**: 2026.5.3
... | closed | completed | false | 1 | [] | [] | 2026-05-04T19:33:12Z | 2026-05-04T22:31:38Z | 2026-05-04T22:31:38Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Lidon-Web | 58,059,636 | MDQ6VXNlcjU4MDU5NjM2 | User | false |
openclaw/openclaw | 4,379,101,478 | I_kwDOQb6kR88AAAABBQPJJg | 77,493 | https://github.com/openclaw/openclaw/issues/77493 | https://api.github.com/repos/openclaw/openclaw/issues/77493 | [Bug]: session_status inside heartbeat reports session/default model instead of active heartbeat.model override | ## Summary
When a heartbeat turn uses `heartbeat.model` as a per-turn override, `session_status({"sessionKey":"current"})` inside that heartbeat can still report the agent/session default model instead of the model actually executing the heartbeat turn.
In my case, the heartbeat is configured to run on `openai-codex/... | open | null | false | 1 | [] | [] | 2026-05-04T19:33:02Z | 2026-05-04T19:41:26Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | vice-magus-faolan | 279,134,512 | U_kgDOEKNBMA | User | false |
openclaw/openclaw | 4,379,142,437 | I_kwDOQb6kR88AAAABBQRpJQ | 77,497 | https://github.com/openclaw/openclaw/issues/77497 | https://api.github.com/repos/openclaw/openclaw/issues/77497 | [Feature]: [cron] Include event timestamp in failure alert body | ### Summary
When a cron job fails, the alert delivered to a configured channel (e.g. Telegram) only contains the job name and error message. Example we got:
⚠️ Cron job "archivist-monitor" failed: FailoverError: LLM request failed: network connection error.
Recipients have to rely on the channel's own message timest... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-05-04T19:39:11Z | 2026-05-04T19:47:34Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Volodymyr120473 | 157,591,057 | U_kgDOCWSmEQ | User | false |
openclaw/openclaw | 4,379,011,890 | I_kwDOQb6kR88AAAABBQJrMg | 77,483 | https://github.com/openclaw/openclaw/issues/77483 | https://api.github.com/repos/openclaw/openclaw/issues/77483 | 2026.4.27 → 2026.5.2 silently drops built-in channel extensions (whatsapp, discord, bluebubbles, ...) — built-in→external @openclaw/* migration without auto-install or warning | ### Bug type
Regression (worked before, now fails) / packaging+migration UX
### Summary
Upgrading OpenClaw from `2026.4.27` to `2026.5.2` (or doing a fresh `2026.5.2` install) leaves all built-in channel extensions silently absent from the active install. The user's pre-existing WhatsApp (and other channels) goes of... | closed | completed | false | 1 | [] | [
"vincentkoc"
] | 2026-05-04T19:16:40Z | 2026-05-04T22:22:17Z | 2026-05-04T22:22:17Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mids-neo | 117,532,760 | U_kgDOBwFoWA | User | false |
openclaw/openclaw | 4,379,195,488 | I_kwDOQb6kR88AAAABBQU4YA | 77,498 | https://github.com/openclaw/openclaw/issues/77498 | https://api.github.com/repos/openclaw/openclaw/issues/77498 | [Bug]: Control UI enters recursive message loop after approving exec commands | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
After approving an exec command in the Control UI webchat, the session enters a recursive processing loop where "Exec finished" completion events are repeatedly re-processed. Only a browser refresh stops it.
This is a distinct variant of t... | closed | completed | false | 1 | [] | [] | 2026-05-04T19:47:07Z | 2026-05-04T21:43:50Z | 2026-05-04T21:43:50Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | joemisra | 326,734 | MDQ6VXNlcjMyNjczNA== | User | false |
openclaw/openclaw | 4,379,026,753 | I_kwDOQb6kR88AAAABBQKlQQ | 77,485 | https://github.com/openclaw/openclaw/issues/77485 | https://api.github.com/repos/openclaw/openclaw/issues/77485 | [Bug]: ReplyRunAlreadyActiveError fires every other gateway-WS chat call, causing 50% reply failure on 2026.5.3 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Every other WebSocket chat request through the gateway fails with ReplyRunAlreadyActiveError, causing 50% reply failure; reproduces deterministically on 2026.5.3 and works fine on 2026.4.26.
### Steps to reproduce
1. Insta... | closed | completed | false | 3 | [
"bug",
"regression"
] | [] | 2026-05-04T19:19:28Z | 2026-05-04T21:05:08Z | 2026-05-04T21:04:56Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | bws14email | 265,138,360 | U_kgDOD82wuA | User | false |
openclaw/openclaw | 4,379,291,183 | I_kwDOQb6kR88AAAABBQauLw | 77,504 | https://github.com/openclaw/openclaw/issues/77504 | https://api.github.com/repos/openclaw/openclaw/issues/77504 | [Bug]: Agent messages disappear in WebChat after brief display — history reload race (repro on #66316 RC#1) | # Bug Report: WebChat Messages Disappear After Brief Display
**Date:** 2026-05-04
**Reporter:** Tim O'Brien (via SmartAlec)
**Severity:** High — breaks usability of primary control surface
**Component:** WebChat UI / Control Dashboard
**OpenClaw Version:** 2026.5.2 (8b2a6e5)
---
## Summary
Agent (assistant)... | closed | completed | false | 1 | [] | [] | 2026-05-04T20:02:55Z | 2026-05-04T20:12:03Z | 2026-05-04T20:12:03Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | legupsystems | 224,921,571 | U_kgDODWgH4w | User | false |
openclaw/openclaw | 4,379,314,303 | I_kwDOQb6kR88AAAABBQcIfw | 77,506 | https://github.com/openclaw/openclaw/issues/77506 | https://api.github.com/repos/openclaw/openclaw/issues/77506 | [Bug]: active-memory toolsAllow references memory_recall but no plugin registers it — causes session wipe | ## Summary
The `active-memory` plugin requests `toolsAllow: ["memory_recall", "memory_search", "memory_get"]` for its embedded sub-agent runs, but no plugin registers a tool named `memory_recall`. The `memory-core` plugin only registers `memory_search` and `memory_get`. This causes the tool allowlist resolver to fail ... | closed | completed | false | 1 | [] | [] | 2026-05-04T20:06:40Z | 2026-05-04T21:47:40Z | 2026-05-04T21:47:40Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | AlexStephenLytton | 5,282,456 | MDQ6VXNlcjUyODI0NTY= | User | false |
openclaw/openclaw | 4,379,348,607 | I_kwDOQb6kR88AAAABBQeOfw | 77,508 | https://github.com/openclaw/openclaw/issues/77508 | https://api.github.com/repos/openclaw/openclaw/issues/77508 | Misleading error when channel block is missing during channels login | `openclaw channels login --channel whatsapp` succeeds and saves credentials, but if `channels.whatsapp` doesn't exist in `openclaw.json`, gateway returns:
```text
Local login saved auth for whatsapp/default, but the running gateway did not restart it: invalid channels.start channel
```
Two problems:
1. `invalid chan... | open | null | false | 1 | [] | [] | 2026-05-04T20:11:50Z | 2026-05-04T20:19:58Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | xooid | 6,255,183 | MDQ6VXNlcjYyNTUxODM= | User | false |
openclaw/openclaw | 4,379,437,793 | I_kwDOQb6kR88AAAABBQjq4Q | 77,517 | https://github.com/openclaw/openclaw/issues/77517 | https://api.github.com/repos/openclaw/openclaw/issues/77517 | [withdrawn] | Withdrawn. | closed | not_planned | false | 1 | [] | [] | 2026-05-04T20:26:47Z | 2026-05-04T20:32:41Z | 2026-05-04T20:32:04Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | jrex-jooni | 266,595,103 | U_kgDOD-PrHw | User | false |
openclaw/openclaw | 4,379,277,250 | I_kwDOQb6kR88AAAABBQZ3wg | 77,503 | https://github.com/openclaw/openclaw/issues/77503 | https://api.github.com/repos/openclaw/openclaw/issues/77503 | [Bug]: iMessage channel polling broken in 2026.5.3-1 (regression from 2026.5.2) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
iMessage channel polling broken in 2026.5.3-1 - messages arrive in database but gateway doesn't respond automatically
### Steps to reproduce
1. Configure iMessage channel in ~/.openclaw/openclaw.json with allowFrom filter... | open | null | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T20:00:24Z | 2026-05-04T20:34:42Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | avavavbot | 281,745,496 | U_kgDOEMsYWA | User | false |
openclaw/openclaw | 4,379,358,395 | I_kwDOQb6kR88AAAABBQe0uw | 77,511 | https://github.com/openclaw/openclaw/issues/77511 | https://api.github.com/repos/openclaw/openclaw/issues/77511 | Move package update execution into an external updater | ## Summary
Follow up on #77282 by moving `openclaw update` execution into an external updater, so the code performing the package replacement is not running from the package being replaced.
## Background
#77282 fixes the immediate stale `dist` failure by adding staged swaps for npm-prefix-shaped installs and compati... | open | null | false | 3 | [
"maintainer"
] | [
"Patrick-Erichsen"
] | 2026-05-04T20:13:18Z | 2026-05-04T21:05:38Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Patrick-Erichsen | 20,157,849 | MDQ6VXNlcjIwMTU3ODQ5 | User | false |
openclaw/openclaw | 4,379,399,645 | I_kwDOQb6kR88AAAABBQhV3Q | 77,514 | https://github.com/openclaw/openclaw/issues/77514 | https://api.github.com/repos/openclaw/openclaw/issues/77514 | [Bug] Telegram forum topic: automatic agent replies not delivered to topic thread (responses silently lost) | ## Summary
Automatic agent responses in Telegram forum/supergroup topic threads are not delivered to the Telegram chat. The agent processes the message and generates a response successfully (visible in session transcript), but the response never appears in the Telegram topic. Manual sends via the `message` tool work c... | closed | not_planned | false | 4 | [] | [] | 2026-05-04T20:19:51Z | 2026-05-04T21:47:31Z | 2026-05-04T21:44:00Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | clawdieclawdita | 259,446,051 | U_kgDOD3bVIw | User | false |
openclaw/openclaw | 4,379,466,006 | I_kwDOQb6kR88AAAABBQlZFg | 77,518 | https://github.com/openclaw/openclaw/issues/77518 | https://api.github.com/repos/openclaw/openclaw/issues/77518 | Webchat persistence regression: embedded-runtime turns skip persistCliTurnTranscript causing silent assistant message drops | ## Summary
Webchat (and other embedded-runtime) sessions silently drop assistant turns from the session JSONL while user turns persist normally. The UI streams the full assistant response, but on the next turn the agent loads the JSONL, sees only user messages, and produces a fresh "first attempt" response — manifesti... | open | null | false | 1 | [] | [] | 2026-05-04T20:31:17Z | 2026-05-04T20:38:56Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | aderius | 78,624,045 | MDQ6VXNlcjc4NjI0MDQ1 | User | false |
openclaw/openclaw | 4,379,470,874 | I_kwDOQb6kR88AAAABBQlsGg | 77,519 | https://github.com/openclaw/openclaw/issues/77519 | https://api.github.com/repos/openclaw/openclaw/issues/77519 | [Bug]: getCurrentPluginMetadataSnapshot workspaceDir mismatch defeats snapshot reuse — 192-stat sweep per dispatch (regression of #76182 / #73353 fix) | ## TL;DR
The fix shipped for #76182 and #73353 added a `getCurrentPluginMetadataSnapshot` reuse path, but the setter (`setCurrentPluginMetadataSnapshot` at `server.impl.ts`) does **not** pass `workspaceDir`, while every reader (e.g. `models-config.ts`, `model-catalog.ts`, `tools.ts`) **does** pass a concrete `workspac... | closed | completed | false | 1 | [] | [] | 2026-05-04T20:31:58Z | 2026-05-05T00:39:36Z | 2026-05-05T00:39:36Z | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | wiipud | 119,115,899 | U_kgDOBxmQew | User | false |
openclaw/openclaw | 4,379,239,968 | I_kwDOQb6kR88AAAABBQXmIA | 77,500 | https://github.com/openclaw/openclaw/issues/77500 | https://api.github.com/repos/openclaw/openclaw/issues/77500 | [Bug]: Gateway spawns 60+ concurrent openclaw processes CPU at 100% on all cores. | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
I have a main agent when I start that agent, 60+ openclaw related processes begin, my CPU is pinned at 100% and nothing can get done on the machine.
### Steps to reproduce
I don't really know - it wasn't doing this before,... | closed | completed | false | 3 | [
"bug",
"regression"
] | [] | 2026-05-04T19:53:56Z | 2026-05-04T20:39:42Z | 2026-05-04T20:18:21Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Kaotic3 | 80,522,225 | MDQ6VXNlcjgwNTIyMjI1 | User | false |
openclaw/openclaw | 4,378,849,673 | I_kwDOQb6kR88AAAABBP_xiQ | 77,471 | https://github.com/openclaw/openclaw/issues/77471 | https://api.github.com/repos/openclaw/openclaw/issues/77471 | [Feature]: Integrate OpenClaw Memory Search with QMD MCP Service | ### Summary
Problem to be Solved
OpenClaw's `memory search` tool adds **~60 seconds latency** on first invocation in every session because it spawns a fresh Python process and loads a 328MB GGUF model from disk each time.
Meanwhile, a **warm QMD MCP service already exists** and can respond in **under 1 second** — bu... | closed | not_planned | false | 1 | [
"enhancement"
] | [] | 2026-05-04T18:48:00Z | 2026-05-04T20:40:21Z | 2026-05-04T20:40:21Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | najef1979-code | 233,555,716 | U_kgDODevHBA | User | false |
openclaw/openclaw | 4,379,348,677 | I_kwDOQb6kR88AAAABBQeOxQ | 77,509 | https://github.com/openclaw/openclaw/issues/77509 | https://api.github.com/repos/openclaw/openclaw/issues/77509 | Persistent compiled runtime output warning for @openclaw/whatsapp despite valid dist output | `@openclaw/whatsapp@2026.5.3` ships `dist/` correctly, verified via `npm pack --dry-run`, but emits this warning on every gateway start:
```text
plugins: plugin: installed plugin package requires compiled runtime output for TypeScript entry index.ts: expected ./dist/index.js, ./dist/index.mjs, ./dist/index.cjs, index.... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T20:11:51Z | 2026-05-04T20:40:28Z | 2026-05-04T20:40:28Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | xooid | 6,255,183 | MDQ6VXNlcjYyNTUxODM= | User | false |
openclaw/openclaw | 4,379,348,749 | I_kwDOQb6kR88AAAABBQePDQ | 77,510 | https://github.com/openclaw/openclaw/issues/77510 | https://api.github.com/repos/openclaw/openclaw/issues/77510 | Document dmPolicy pairing interaction with allowFrom for WhatsApp DMs | The default `dmPolicy: "pairing"` silently requires a separate pairing step on top of `allowFrom`.
Observed behaviour:
- a phone number is present in `allowFrom`
- WhatsApp channel is configured
- direct messages from that number are still dropped if the number has not also been paired
The docs at `/channels/whatsap... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T20:11:51Z | 2026-05-04T20:40:34Z | 2026-05-04T20:40:34Z | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | xooid | 6,255,183 | MDQ6VXNlcjYyNTUxODM= | User | false |
openclaw/openclaw | 4,379,472,823 | I_kwDOQb6kR88AAAABBQlztw | 77,520 | https://github.com/openclaw/openclaw/issues/77520 | https://api.github.com/repos/openclaw/openclaw/issues/77520 | [Bug]: Discord channel turn marks done but drops final reply despite visibleReplies automatic | ## Summary
Discord guild/channel turns can complete and show the terminal ✅ status reaction, but no final assistant text is delivered to Discord even when `messages.groupChat.visibleReplies` is explicitly set to `automatic`.
This looks related to the 2026.4.27+ group/channel reply-mode changes, but it is not fixed by... | closed | completed | false | 1 | [] | [] | 2026-05-04T20:32:18Z | 2026-05-05T00:33:23Z | 2026-05-05T00:33:23Z | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | jsunderulo | 193,991,514 | U_kgDOC5ATWg | User | false |
openclaw/openclaw | 4,379,597,380 | I_kwDOQb6kR88AAAABBQtaRA | 77,525 | https://github.com/openclaw/openclaw/issues/77525 | https://api.github.com/repos/openclaw/openclaw/issues/77525 | /status shows ?/1.0m for xAI Grok models instead of actual context usage | ### Description
When using `xai/grok-4.3` (or other Grok variants), the status display shows `Context: ?/1.0m` instead of real usage numbers, while all other providers (Anthropic, OpenAI, Google) correctly show `28k/200k (14%)` style output.
### Reproduction
1. Run a session with primary model `xai/grok-4.3`
2. Run `/... | open | null | false | 4 | [] | [] | 2026-05-04T20:53:11Z | 2026-05-04T22:45:06Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | PrinceOfEgypt | 2,649,235 | MDQ6VXNlcjI2NDkyMzU= | User | false |
openclaw/openclaw | 4,379,608,404 | I_kwDOQb6kR88AAAABBQuFVA | 77,526 | https://github.com/openclaw/openclaw/issues/77526 | https://api.github.com/repos/openclaw/openclaw/issues/77526 | Discord channel: gateway WebSocket never reaches `ready` on 2026.5.2 and 2026.5.3 — inbound `messageCreate` events never dispatch | ## Summary
After updating from `openclaw@2026.3.24` to `2026.5.2` (and confirmed still broken on `2026.5.3`), the Discord channel never finishes its WebSocket gateway handshake. The bot logs `discord client initialized as <id>; awaiting gateway readiness` and then nothing further. No `messageCreate` events ever fire, ... | closed | completed | false | 2 | [] | [] | 2026-05-04T20:55:03Z | 2026-05-04T22:54:23Z | 2026-05-04T22:54:23Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | sacred8-Dev | 281,758,430 | U_kgDOEMtK3g | User | false |
openclaw/openclaw | 4,379,682,883 | I_kwDOQb6kR88AAAABBQyoQw | 77,532 | https://github.com/openclaw/openclaw/issues/77532 | https://api.github.com/repos/openclaw/openclaw/issues/77532 | Agent bootstrap takes 3+ minutes per turn; core-plugin-tools, system-prompt, stream-setup each 45-75s | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Every agent turn takes 3-5 minutes to start streaming on a clean install; trace logs show core-plugin-tools (~73s), system-prompt (~45s), and stream-setup (~45s) as the dominant costs on every turn, unaffected ... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-05-04T21:08:32Z | 2026-05-05T09:53:33Z | null | NONE | null | 20260505T120033Z | 2026-05-05T12:00:33Z | Maxtonairay | 275,559,522 | U_kgDOEGy0Yg | User | false |
openclaw/openclaw | 4,379,676,192 | I_kwDOQb6kR88AAAABBQyOIA | 77,531 | https://github.com/openclaw/openclaw/issues/77531 | https://api.github.com/repos/openclaw/openclaw/issues/77531 | Regression in 2026.5.3-1: CLI transcript persistence fails with "Context engine 'legacy' is not registered" | ## Summary
After upgrading to `2026.5.3-1`, the CLI transcript persistence path again throws
```
CLI transcript persistence failed ... Context engine "legacy" is not registered
```
This was previously reported and fixed under #75546 (closed 2026-05-01) and the related `/subagents` regression #74865. The same symptom... | closed | completed | false | 2 | [] | [] | 2026-05-04T21:07:17Z | 2026-05-04T21:20:52Z | 2026-05-04T21:20:52Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | nikolaykazakovvs-ux | 233,889,353 | U_kgDODfDeSQ | User | false |
openclaw/openclaw | 4,379,737,698 | I_kwDOQb6kR88AAAABBQ1-Yg | 77,534 | https://github.com/openclaw/openclaw/issues/77534 | https://api.github.com/repos/openclaw/openclaw/issues/77534 | writeSavedMediaBuffer fsyncs a read-only file handle → EPERM on Windows (drops all media uploads) | ## Summary
On Windows, every Telegram/Discord image upload fails with:
```
EPERM: operation not permitted, fsync
at FileHandle.sync (node:internal/fs/promises:...)
at writeSavedMediaBuffer (dist/store-<hash>.js)
```
The runtime never delivers the file to the agent. Reproduced on OpenClaw `2026.5.3-1`, Node `22.2... | closed | completed | false | 1 | [] | [] | 2026-05-04T21:19:07Z | 2026-05-04T21:28:07Z | 2026-05-04T21:28:07Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | joeywrightphoto | 114,463,115 | U_kgDOBtKRiw | User | false |
openclaw/openclaw | 4,379,671,077 | I_kwDOQb6kR88AAAABBQx6JQ | 77,530 | https://github.com/openclaw/openclaw/issues/77530 | https://api.github.com/repos/openclaw/openclaw/issues/77530 | Can't update to new main - npm errors | I tried updating to the new main again today since I haven't heard anything on Discord or on here (I see that it's closed). Here's the console output:
peter@SkytechRGB3D:~$ openclaw update --tag main
Updating OpenClaw...
Stopping managed gateway service before package update...
Stopped systemd service: openclaw-gatew... | closed | completed | false | 2 | [] | [] | 2026-05-04T21:06:18Z | 2026-05-04T21:35:24Z | 2026-05-04T21:35:24Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | PeterTremonti | 56,261,993 | MDQ6VXNlcjU2MjYxOTkz | User | false |
openclaw/openclaw | 4,379,815,491 | I_kwDOQb6kR88AAAABBQ6uQw | 77,541 | https://github.com/openclaw/openclaw/issues/77541 | https://api.github.com/repos/openclaw/openclaw/issues/77541 | Feature request: per-channel restart RPC for surgical channel recycling | ## Problem
There is currently no way to restart a single misbehaving channel in OpenClaw. The only programmatic restart available is `systemctl --user restart openclaw-gateway`, which is too disruptive — it kills **all** channels, **all** crons, and **all** active sessions just to recycle one flapping connection.
For... | closed | not_planned | false | 1 | [] | [] | 2026-05-04T21:32:54Z | 2026-05-04T21:39:51Z | 2026-05-04T21:39:51Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jared-rebel | 147,876,109 | U_kgDOCNBpDQ | User | false |
openclaw/openclaw | 4,379,857,212 | I_kwDOQb6kR88AAAABBQ9RPA | 77,548 | https://github.com/openclaw/openclaw/issues/77548 | https://api.github.com/repos/openclaw/openclaw/issues/77548 | Telegram DM session not routing inbound messages to agent | ## Problem
Inbound Telegram direct messages to @AbdulClawBot are not reaching the agent. Messages arrive at Telegram's servers, OpenClaw webhook endpoint is called, but the agent never responds. Outbound DMs from the agent work fine. Topic/thread routing (e.g. threadId=959) works fine. Only inbound DM routing appears ... | open | null | false | 1 | [] | [] | 2026-05-04T21:40:14Z | 2026-05-05T00:17:36Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | tasi333 | 7,864,516 | MDQ6VXNlcjc4NjQ1MTY= | User | false |
openclaw/openclaw | 4,379,856,346 | I_kwDOQb6kR88AAAABBQ9N2g | 77,546 | https://github.com/openclaw/openclaw/issues/77546 | https://api.github.com/repos/openclaw/openclaw/issues/77546 | [Feature]: Alert and audit unauthorized Telegram DM/group access attempts | ### Summary
Add first-class security visibility for unauthorized Telegram inbound attempts that are blocked by `dmPolicy`, `groupPolicy`, `allowFrom`, `groupAllowFrom`, or `groups` allowlists.
### Problem to solve
OpenClaw can correctly block unauthorized Telegram DMs and non-allowlisted group contexts, but operator... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-05-04T21:40:08Z | 2026-05-04T21:48:12Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,379,903,910 | I_kwDOQb6kR88AAAABBRAHpg | 77,551 | https://github.com/openclaw/openclaw/issues/77551 | https://api.github.com/repos/openclaw/openclaw/issues/77551 | Bedrock aws-sdk auth: ExpiredTokenException after credential file refresh (regression in v2026.5.3) | ## Description
After upgrading to v2026.5.3-1, the gateway no longer picks up refreshed AWS credentials from `~/.aws/credentials` during its lifetime. Every credential rotation (e.g. `mwinit` / `ada credentials update`) causes `ExpiredTokenException` until the gateway is manually restarted.
This did **not** happen on... | open | null | false | 1 | [] | [] | 2026-05-04T21:49:31Z | 2026-05-04T21:58:22Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | shipitsteven | 62,415,298 | MDQ6VXNlcjYyNDE1Mjk4 | User | false |
openclaw/openclaw | 4,380,000,675 | I_kwDOQb6kR88AAAABBRGBow | 77,561 | https://github.com/openclaw/openclaw/issues/77561 | https://api.github.com/repos/openclaw/openclaw/issues/77561 | ownsCompaction tool-result context guard bypass still present in v2026.5.3 (Codex sessions have no mid-turn compaction protection) | Follow-up to #71325 (closed/locked April 25).
**The `ownsCompaction` bypass is still present in v2026.5.3 (06d46f7).** At `selection-BgogoK8g.js:7163`, the tool-result context guard is only installed when `ownsCompaction !== true`:
```javascript
if (!activeContextEngine || activeContextEngine.info.ownsCompaction !== ... | open | null | false | 1 | [] | [] | 2026-05-04T22:08:19Z | 2026-05-04T22:16:54Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | KelTech-Services | 18,705,530 | MDQ6VXNlcjE4NzA1NTMw | User | false |
openclaw/openclaw | 4,380,052,553 | I_kwDOQb6kR88AAAABBRJMSQ | 77,562 | https://github.com/openclaw/openclaw/issues/77562 | https://api.github.com/repos/openclaw/openclaw/issues/77562 | [Bug]: Control UI /new creates empty session with no hook firing — session-memory broken, custom hooks broken, /new feels completely broken | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After v2026.5.2, typing `/new` in the Control UI (webchat) no longer fires `command:new` internal hooks because commit `37aebf612b` rerouted it from `sendChatMessageNow(host, "/new")` (which went through Gateway command disp... | closed | not_planned | false | 2 | [] | [] | 2026-05-04T22:18:43Z | 2026-05-04T22:31:47Z | 2026-05-04T22:31:47Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | php737 | 31,496,660 | MDQ6VXNlcjMxNDk2NjYw | User | false |
openclaw/openclaw | 4,380,095,151 | I_kwDOQb6kR88AAAABBRLyrw | 77,563 | https://github.com/openclaw/openclaw/issues/77563 | https://api.github.com/repos/openclaw/openclaw/issues/77563 | [Bug]: release/2026.5.3, feishu plugin group message no reply | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
no error, openclaw-gateway-1 | 2026-05-05T06:13:47.745+08:00 [feishu] feishu[main]: dispatch complete (queuedFinal=false, replies=0)
### Steps to reproduce
~
### Expected behavior
reply group message
### Actual behavio... | closed | not_planned | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T22:28:46Z | 2026-05-04T22:39:38Z | 2026-05-04T22:39:38Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | havcino | 50,948,638 | MDQ6VXNlcjUwOTQ4NjM4 | User | false |
openclaw/openclaw | 4,380,110,930 | I_kwDOQb6kR88AAAABBRMwUg | 77,566 | https://github.com/openclaw/openclaw/issues/77566 | https://api.github.com/repos/openclaw/openclaw/issues/77566 | [Bug]: Gemini 3.x silent hang on subagent flows — `thoughtSignature` dropped on cross-provider replay (2026.5.3-1) | ## Summary
`google/gemini-3.1-pro-preview` (`thinkingLevel: high`) reliably hangs after a few tool-call rounds in any flow where conversation history was authored by a different provider/model — e.g. subagents spawned from a Claude or GPT orchestrator. Failure mode: gateway logs `LLM idle timeout (600s): no response f... | open | null | false | 3 | [] | [] | 2026-05-04T22:32:23Z | 2026-05-05T05:57:00Z | null | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | jrex-jooni | 266,595,103 | U_kgDOD-PrHw | User | false |
openclaw/openclaw | 4,380,144,862 | I_kwDOQb6kR88AAAABBRO03g | 77,567 | https://github.com/openclaw/openclaw/issues/77567 | https://api.github.com/repos/openclaw/openclaw/issues/77567 | Feature request: Uncle Jim mode for isolated family agents | # Feature request: “Uncle Jim mode” for isolated family agents
## Summary
Add a first-class operating mode for isolated agents that support non-technical family members. The mode should let an agent help normally, but automatically surface agenda items, errors, uncertainty, and missing technical capabilities/setup req... | open | null | false | 1 | [] | [] | 2026-05-04T22:40:13Z | 2026-05-04T22:47:21Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,380,241,308 | I_kwDOQb6kR88AAAABBRUtnA | 77,570 | https://github.com/openclaw/openclaw/issues/77570 | https://api.github.com/repos/openclaw/openclaw/issues/77570 | [Bug]: OpenAI Responses subagent flows hang on phantom `wss://api.openai.com/v1/responses` URL (2026.5.3-1; #77070 autoclosed without fix) | ## Summary
OpenAI Responses-API multi-turn tool-call flows reliably hang in shipped 2026.5.3-1 with `[llm-idle-timeout] openai/<model> produced no reply before the idle watchdog`. Trivial single-turn calls to the same models with the same key complete in ~1–2 seconds direct to the OpenAI HTTP/SSE endpoint. The shipped... | closed | completed | false | 2 | [] | [
"vincentkoc"
] | 2026-05-04T23:01:21Z | 2026-05-04T23:07:08Z | 2026-05-04T23:07:08Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | jrex-jooni | 266,595,103 | U_kgDOD-PrHw | User | false |
openclaw/openclaw | 4,380,251,106 | I_kwDOQb6kR88AAAABBRVT4g | 77,571 | https://github.com/openclaw/openclaw/issues/77571 | https://api.github.com/repos/openclaw/openclaw/issues/77571 | [Feature]: Optionally inject extraDirs skills into the <available_skills> prompt block | ### Summary
Skills loaded from skills.load.extraDirs are discovered and marked ✓ ready by the gateway, but are not included in the <available_skills> block injected into the agent's system prompt. This prevents agents from auto-selecting them during the pre-reply skill scan.
### Problem to solve
In a multi-agent dep... | closed | completed | false | 1 | [
"enhancement"
] | [] | 2026-05-04T23:03:38Z | 2026-05-04T23:11:31Z | 2026-05-04T23:11:24Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | elyalvarado | 545,352 | MDQ6VXNlcjU0NTM1Mg== | User | false |
openclaw/openclaw | 4,380,278,544 | I_kwDOQb6kR88AAAABBRW_EA | 77,574 | https://github.com/openclaw/openclaw/issues/77574 | https://api.github.com/repos/openclaw/openclaw/issues/77574 | [Bug]: clawdock-dashboard switches the running gateway image | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After running clawdock-dashboard --no-open, docker ps showed openclaw-openclaw-gateway-1 using ghcr.io/openclaw/openclaw:latest instead of the previously observed openclaw-gateway:tools.
### Steps to reproduce
1. On Ubuntu... | open | null | false | 1 | [
"bug",
"regression"
] | [] | 2026-05-04T23:10:16Z | 2026-05-05T00:39:43Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | mattharvill | 1,859,540 | MDQ6VXNlcjE4NTk1NDA= | User | false |
openclaw/openclaw | 4,380,290,037 | I_kwDOQb6kR88AAAABBRXr9Q | 77,576 | https://github.com/openclaw/openclaw/issues/77576 | https://api.github.com/repos/openclaw/openclaw/issues/77576 | [Regression] Telegram group session responses route to webchat instead of back to Telegram (v2026.5.3-1) | ## Summary
When a Telegram group message triggers an agent session (via `channels.telegram.groups` binding), the agent generates a correct response but the response is delivered to the **webchat Control UI** session instead of back to the **Telegram group**. This is a regression — explicit `message` tool sends work co... | open | null | false | 1 | [] | [] | 2026-05-04T23:12:46Z | 2026-05-04T23:20:56Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Childrenic | 263,755,971 | U_kgDOD7iYww | User | false |
openclaw/openclaw | 4,380,302,557 | I_kwDOQb6kR88AAAABBRYc3Q | 77,577 | https://github.com/openclaw/openclaw/issues/77577 | https://api.github.com/repos/openclaw/openclaw/issues/77577 | [Bug]: Token-in-URL works on / but not on /__openclaw__/canvas/ | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
With the same gateway token in the query string, http://127.0.0.1:18789/__openclaw__/canvas/?token=… did not work over the Mac SSH tunnel, while http://127.0.0.1:18789/?token=… did.
### Steps to reproduce
1. ... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-05-04T23:15:32Z | 2026-05-04T23:23:54Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mattharvill | 1,859,540 | MDQ6VXNlcjE4NTk1NDA= | User | false |
openclaw/openclaw | 4,380,369,974 | I_kwDOQb6kR88AAAABBRckNg | 77,579 | https://github.com/openclaw/openclaw/issues/77579 | https://api.github.com/repos/openclaw/openclaw/issues/77579 | [Bug]: Telegram polling stalls indefinitely — gateway stops receiving messages until manual restart | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
## Summary
The Telegram channel polling (`getUpdates`) stalls indefinitely and never ... | closed | not_planned | false | 1 | [
"bug",
"bug:crash"
] | [] | 2026-05-04T23:30:25Z | 2026-05-04T23:39:01Z | 2026-05-04T23:39:01Z | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | mdoroudi | 687,773 | MDQ6VXNlcjY4Nzc3Mw== | User | false |
openclaw/openclaw | 4,380,396,107 | I_kwDOQb6kR88AAAABBReKSw | 77,582 | https://github.com/openclaw/openclaw/issues/77582 | https://api.github.com/repos/openclaw/openclaw/issues/77582 | [Bug]: Telegram selected quote context is normalized but omitted from final model prompt | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Telegram selected quote context is captured in OpenClaw’s normalized runtime context, but it is not included in the final model prompt / finalPromptText. This causes short Telegram replies like “explain,” “why,” “fix this,” ... | closed | completed | false | 2 | [
"bug",
"regression"
] | [] | 2026-05-04T23:35:34Z | 2026-05-05T07:38:31Z | 2026-05-05T03:20:08Z | NONE | null | 20260505T090045Z | 2026-05-05T09:00:45Z | Nickfost | 1,572,453 | MDQ6VXNlcjE1NzI0NTM= | User | false |
openclaw/openclaw | 4,380,394,396 | I_kwDOQb6kR88AAAABBReDnA | 77,581 | https://github.com/openclaw/openclaw/issues/77581 | https://api.github.com/repos/openclaw/openclaw/issues/77581 | Control UI thinking dropdown shows `Default (off)` when effective inherited thinking is high | ## Summary
The Control UI chat/session thinking dropdown displays `Default (off)` when a session has no per-session thinking override, even though it is effectively inheriting a configured default such as `agents.defaults.thinkingDefault: "high"`.
Runtime behavior appears correct. This is a UI labeling bug: the label... | open | null | false | 1 | [] | [] | 2026-05-04T23:35:17Z | 2026-05-04T23:40:43Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | prasith-jobleap | 181,791,026 | U_kgDOCtXpMg | User | false |
openclaw/openclaw | 4,380,486,489 | I_kwDOQb6kR88AAAABBRjrWQ | 77,586 | https://github.com/openclaw/openclaw/issues/77586 | https://api.github.com/repos/openclaw/openclaw/issues/77586 | Commitment extraction fails when tools.alsoAllow is set (disableTools: true doesn't bypass allowlist) | ## Bug
Commitment extraction runs with `disableTools: true` (line ~379 in `runtime-DXR5S3Pt.js`, `defaultExtractBatch`), but `runEmbeddedPiAgent` still resolves the tool allowlist before checking `disableTools`. When `tools.alsoAllow` contains tool names (e.g., `["llm-task", "exec"]`), the resolved allowlist includes ... | closed | completed | false | 1 | [] | [] | 2026-05-04T23:57:25Z | 2026-05-05T00:06:28Z | 2026-05-05T00:06:21Z | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | Orionation | 260,579,663 | U_kgDOD4ghTw | User | false |
openclaw/openclaw | 4,380,549,358 | I_kwDOQb6kR88AAAABBRng7g | 77,594 | https://github.com/openclaw/openclaw/issues/77594 | https://api.github.com/repos/openclaw/openclaw/issues/77594 | [Feature]: Discord: missed DM messages during gateway restart (no catchup on reconnect) | ### Summary
When the OpenClaw gateway restarts, any Discord DM messages sent while it was offline
are permanently lost. The bot has no mechanism to recover them after reconnecting.
This is particularly painful for DMs because:
- Guild channel messages can at least be inferred from visible channel state
- DMs have no... | closed | not_planned | false | 1 | [
"enhancement"
] | [] | 2026-05-05T00:13:12Z | 2026-05-05T00:21:32Z | 2026-05-05T00:21:25Z | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | sergattic | 36,697,733 | MDQ6VXNlcjM2Njk3NzMz | User | false |
openclaw/openclaw | 4,380,598,588 | I_kwDOQb6kR88AAAABBRqhPA | 77,600 | https://github.com/openclaw/openclaw/issues/77600 | https://api.github.com/repos/openclaw/openclaw/issues/77600 | [Document correct] Clarify Codex backend OAuth setup for isolated agents | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Clarify the correct OAuth setup for isolated OpenClaw agents using the native Codex backend.
The intended route is:
`OpenClaw isolated agent -> openai/gpt-5.5 -> codex backend/runtime -> Codex app-server ... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-05-05T00:25:49Z | 2026-05-05T00:34:03Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | wingleungron | 281,655,117 | U_kgDOEMm3TQ | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.