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 | 4,113,895,278 | I_kwDOQb6kR871NQ9u | 51,881 | https://github.com/openclaw/openclaw/issues/51881 | https://api.github.com/repos/openclaw/openclaw/issues/51881 | [Bug]: MIME sniff maps PNG files to image/apng, causing Claude API rejection (400 ValidationException) | ## Description
When sending PNG images through the gateway, `fileTypeFromBuffer` (from the `file-type` npm package) sniffs certain PNG files as `image/apng` (Animated PNG). The `parseMessageWithAttachments` function in `chat-attachments.ts` then uses this sniffed MIME over the client-provided MIME, resulting in Claude... | open | null | false | 3 | [] | [] | 2026-03-21T21:23:28Z | 2026-05-04T20:59:36Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Antsoldier1974 | 248,194,488 | U_kgDODssluA | User | false |
openclaw/openclaw | 4,114,003,242 | I_kwDOQb6kR871NrUq | 51,892 | https://github.com/openclaw/openclaw/issues/51892 | https://api.github.com/repos/openclaw/openclaw/issues/51892 | iMessage: pass reply_to_guid to imsg for native threaded replies | ## Summary
When replying to an iMessage, OpenClaw correctly tracks the `replyToId` (message GUID) and prepends a `[[reply_to:<guid>]]` tag to the outbound message text via `prependReplyTagIfNeeded()`. However, this tag is **not extracted and passed as `reply_to_guid`** in the params sent to the `imsg` RPC `send` call.... | open | null | false | 2 | [] | [] | 2026-03-21T22:11:13Z | 2026-05-04T20:59:38Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | rancur | 235,745,911 | U_kgDODg0ydw | User | false |
openclaw/openclaw | 4,114,066,196 | I_kwDOQb6kR871N6sU | 51,903 | https://github.com/openclaw/openclaw/issues/51903 | https://api.github.com/repos/openclaw/openclaw/issues/51903 | bug(agents): nodes tool describe action fails with 'node required' due to schema/implementation mismatch | ### Description
The `nodes` tool fails when an agent calls `action: 'describe'` without a `node` parameter, even though the tool's schema defines `node` as optional. This causes a recurring error cycle where agents attempt discovery via `describe` before listing nodes.
### Root Cause
1. **Schema Mismatch**: In `src... | open | null | false | 4 | [] | [] | 2026-03-21T22:39:12Z | 2026-05-04T20:15:58Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Liuhaai | 55,118,568 | MDQ6VXNlcjU1MTE4NTY4 | User | false |
openclaw/openclaw | 4,114,110,290 | I_kwDOQb6kR871OFdS | 51,911 | https://github.com/openclaw/openclaw/issues/51911 | https://api.github.com/repos/openclaw/openclaw/issues/51911 | Anthropic setup-token onboarding path has multiple failure modes (docs gap, credential propagation, multi-agent sync) | ### Bug type
Regression (worked before, now fails)
### Summary
The Anthropic `setup-token` onboarding path has multiple overlapping failure modes that cause it to appear successful while leaving the system in a broken state. Based on direct experience and review of ~10 related issues, there are three distinct proble... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-03-21T22:59:21Z | 2026-05-04T20:59:40Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ragesaq | 11,304,287 | MDQ6VXNlcjExMzA0Mjg3 | User | false |
openclaw/openclaw | 4,114,136,686 | I_kwDOQb6kR871OL5u | 51,918 | https://github.com/openclaw/openclaw/issues/51918 | https://api.github.com/repos/openclaw/openclaw/issues/51918 | Feature Request: Owner message priority / preemption during agent turns | ## Problem
When the owner sends a message during an active agent turn (with tool calls in execution), the message is queued in FIFO and only delivered when the turn completes.
This causes delays of **30s to 8min** depending on turn complexity.
In critical scenarios (`"para tudo"`, `"cancela"`, `"stop"`), the delay c... | closed | completed | true | 4 | [] | [] | 2026-03-21T23:10:01Z | 2026-05-03T04:34:26Z | 2026-05-01T01:26:57Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | victorbot-cmd | 269,232,838 | U_kgDOEAwqxg | User | false |
openclaw/openclaw | 4,114,350,392 | I_kwDOQb6kR871PAE4 | 51,977 | https://github.com/openclaw/openclaw/issues/51977 | https://api.github.com/repos/openclaw/openclaw/issues/51977 | [Bug]: skillsSnapshot.prompt retention/duplication causes gateway heap growth, OOMs, and hangs | ## Summary
OpenClaw appears to retain and duplicate large rendered skills prompt strings (`skillsSnapshot.prompt`) across session state, especially in isolated cron flows. Over time this drives heap growth to ~4.1 GB, causing either fatal Node/V8 OOM crashes or gateway hangs under near-heap-limit pressure.
Heap snaps... | closed | not_planned | true | 5 | [] | [] | 2026-03-22T01:18:12Z | 2026-05-02T04:28:48Z | 2026-04-29T09:05:48Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | sam-reed-ubind | 261,396,112 | U_kgDOD5SWkA | User | false |
openclaw/openclaw | 4,114,474,681 | I_kwDOQb6kR871Pea5 | 52,029 | https://github.com/openclaw/openclaw/issues/52029 | https://api.github.com/repos/openclaw/openclaw/issues/52029 | Feature Request: heartbeat.tools option to disable tools during heartbeat | # Feature Request: `heartbeat.tools` option to disable tools during heartbeat
## Problem
When using local/small models (3-9B parameters) for heartbeat, models ignore the "do not use any tools" instruction in the prompt and make tool calls anyway. This causes:
1. **Dangerous exec calls** — e.g., 4b model ran `opencla... | open | null | false | 3 | [] | [] | 2026-03-22T02:48:29Z | 2026-05-04T21:17:58Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | andyk-ms | 91,510,251 | U_kgDOBXRV6w | User | false |
openclaw/openclaw | 4,114,555,999 | I_kwDOQb6kR871PyRf | 52,045 | https://github.com/openclaw/openclaw/issues/52045 | https://api.github.com/repos/openclaw/openclaw/issues/52045 | [Bug]: Enable browser hosting and check the browser status; it indicates that the gateway is not enabled | ### Bug type
Regression (worked before, now fails)
### Summary
I used the command openclaw gateway status and it showed the gateway was normal, but when I ran openclaw browser --browser-profile openclaw status, it indicated that the gateway was closed.
<img width="1356" height="900" alt="Image" src="https://github.c... | open | null | false | 3 | [
"bug",
"regression"
] | [] | 2026-03-22T03:28:40Z | 2026-05-04T20:23:20Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ylx1314 | 109,227,716 | U_kgDOBoKuxA | User | false |
openclaw/openclaw | 4,114,561,498 | I_kwDOQb6kR871Pzna | 52,046 | https://github.com/openclaw/openclaw/issues/52046 | https://api.github.com/repos/openclaw/openclaw/issues/52046 | [Feature]: Non-blocking heartbeat execution to prevent message processing delays | ### Summary
Enable heartbeat checks to run without blocking message processing, preventing 5-10 second delays when users send messages during heartbeat execution.
### Problem to solve
Current behavior:
Heartbeat checks (Token monitoring, health checks, security audits) block the main message processing thread for 5-... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-22T03:30:27Z | 2026-05-04T19:11:36Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ccff0 | 265,008,116 | U_kgDOD8uz9A | User | false |
openclaw/openclaw | 4,114,693,637 | I_kwDOQb6kR871QT4F | 52,073 | https://github.com/openclaw/openclaw/issues/52073 | https://api.github.com/repos/openclaw/openclaw/issues/52073 | [Bug]: Agent becomes completely unresponsive during Skill installation and does not report task completion | ### Bug type
Crash (process/app exits or hangs)
### Summary
When instructed via Telegram to install a Skill, the agent goes completely silent — no acknowledgment, no progress update, and no completion notification after the task finishes.
### Steps to reproduce
1. Send a Skill installation instruction to the agent... | open | null | false | 6 | [
"bug",
"bug:crash"
] | [] | 2026-03-22T04:24:38Z | 2026-05-04T19:11:37Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | benbenzhuyi | 126,326,162 | U_kgDOB4eVkg | User | false |
openclaw/openclaw | 4,114,830,565 | I_kwDOQb6kR871Q1Tl | 52,092 | https://github.com/openclaw/openclaw/issues/52092 | https://api.github.com/repos/openclaw/openclaw/issues/52092 | Gateway OOM: Node.js exceeds default V8 heap limit with cron jobs | ## Description
OpenClaw gateway process silently exits when processing cron jobs, exceeding V8's default ~1.5GB heap limit. The process vanishes without a crash log, SIGTERM, or OOM message — systemd just sees it disappear and restarts it via `Restart=always`.
## Environment
- **OpenClaw version:** 2026.3.13 (61d171... | open | null | false | 3 | [] | [] | 2026-03-22T05:05:29Z | 2026-05-04T20:16:02Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Bricklink | 147,354,010 | U_kgDOCMhxmg | User | false |
openclaw/openclaw | 4,115,024,686 | I_kwDOQb6kR871Rksu | 52,105 | https://github.com/openclaw/openclaw/issues/52105 | https://api.github.com/repos/openclaw/openclaw/issues/52105 | [Bug]: Control UI /think command crashes | ### Bug type
Crash (process/app exits or hangs)
### Summary
The command is "swallowed" - it disappears from the input without any response. Error only appears in browser console (F12).
Other slash commands like `/model` work fine.
## Environment
- OpenClaw: 2026.3.13
- Browser: Chrome latest
- OS: Windows
##... | closed | completed | true | 4 | [
"bug",
"bug:crash"
] | [] | 2026-03-22T05:56:18Z | 2026-05-05T04:27:47Z | 2026-05-02T10:08:06Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | MooreQiao | 189,860,296 | U_kgDOC1EJyA | User | false |
openclaw/openclaw | 4,115,193,347 | I_kwDOQb6kR871SN4D | 52,130 | https://github.com/openclaw/openclaw/issues/52130 | https://api.github.com/repos/openclaw/openclaw/issues/52130 | [Bug]: restart storm from telegram.retry.jitter type mismatch + misleading doctor SecretRef for Telegram token | ## Summary
After host reboot, OpenClaw appeared "hung". Investigation showed a restart storm in a secondary runtime, plus confusing SecretRef diagnostics.
Observed behavior:
- `openclaw doctor` reports:
- `channels.telegram.botToken: unresolved SecretRef "file:filemain:/providers/channels/telegram/botToken"`
- At th... | open | null | false | 5 | [] | [] | 2026-03-22T07:14:54Z | 2026-05-04T19:31:02Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | skbotoc1-web | 267,329,629 | U_kgDOD-8gXQ | User | false |
openclaw/openclaw | 4,115,322,537 | I_kwDOQb6kR871Stap | 52,146 | https://github.com/openclaw/openclaw/issues/52146 | https://api.github.com/repos/openclaw/openclaw/issues/52146 | Inject sibling agent list into session system prompt | ## Problem
In a multi-agent setup, each agent's session has no awareness of other agents configured in `agents.list`. The only way for an agent to know about its siblings is through manually maintained documentation (e.g., a table in TOOLS.md).
This leads to:
- Stale information when agents are added/removed but docs... | open | null | false | 2 | [] | [] | 2026-03-22T08:22:04Z | 2026-05-04T20:59:58Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | jxdbaby | 249,119,425 | U_kgDODtlCwQ | User | false |
openclaw/openclaw | 4,115,327,379 | I_kwDOQb6kR871SumT | 52,147 | https://github.com/openclaw/openclaw/issues/52147 | https://api.github.com/repos/openclaw/openclaw/issues/52147 | Agent run timeout during tool execution misclassified as LLM timeout, triggers unnecessary model fallback |
## Bug type
Agent behavior
## Summary
Agent run timeout during long tool execution (e.g. `process(poll)`) is misclassified as "LLM request timed out", triggering unnecessary model fallback — even though the primary model responded correctly.
## Steps to reproduce
1. Configure an agent with `claude-opus-4-6` as prim... | closed | completed | true | 4 | [] | [] | 2026-03-22T08:24:12Z | 2026-05-05T04:27:08Z | 2026-05-02T13:53:19Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | andychu666 | 8,448,251 | MDQ6VXNlcjg0NDgyNTE= | User | false |
openclaw/openclaw | 4,115,436,528 | I_kwDOQb6kR871TJPw | 52,161 | https://github.com/openclaw/openclaw/issues/52161 | https://api.github.com/repos/openclaw/openclaw/issues/52161 | [Bug]: It froze after the search | ### Bug type
Regression (worked before, now fails)
### Summary
It froze after the search
<img width="2550" height="1467" alt="Image" src="https://github.com/user-attachments/assets/69773ee2-3859-4018-bf62-da72f501fc95" />
### Steps to reproduce
When I run `openclaw-tavily` to search for https://github.com/cft0808... | open | null | false | 1 | [
"bug",
"regression"
] | [] | 2026-03-22T09:18:43Z | 2026-05-04T20:16:06Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | MackJack023 | 141,124,084 | U_kgDOCGlh9A | User | false |
openclaw/openclaw | 4,115,576,200 | I_kwDOQb6kR871TrWI | 52,184 | https://github.com/openclaw/openclaw/issues/52184 | https://api.github.com/repos/openclaw/openclaw/issues/52184 | [Feature]: Prefer Volta shim path over version-pinned Volta node path for macOS gateway LaunchAgent | ## Summary
On macOS, `openclaw gateway install` / service generation can pin the gateway LaunchAgent to a version-specific Volta Node path such as:
`/Users/<user>/.volta/tools/image/node/24.14.0/bin/node`
For users who intentionally manage Node with Volta, this is more brittle than necessary. A better choice would b... | open | null | false | 2 | [] | [] | 2026-03-22T10:15:55Z | 2026-05-04T21:00:02Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | real-jacket | 36,723,264 | MDQ6VXNlcjM2NzIzMjY0 | User | false |
openclaw/openclaw | 4,115,883,587 | I_kwDOQb6kR871U2ZD | 52,238 | https://github.com/openclaw/openclaw/issues/52238 | https://api.github.com/repos/openclaw/openclaw/issues/52238 | 飞书话题群(Topic Group)被识别为p2p导致消息路由错误 | ## 问题描述
飞书话题群中的消息被 OpenClaw 识别为 `p2p`（单聊）而非 `group`（群聊），导致消息被错误路由到网页端主会话而非飞书群。
## 复现步骤
1. 将机器人加入飞书话题群
2. 在话题群中向机器人发送消息
3. 观察：消息没有在话题群中回复，而是出现在了网页端
## 日志证据
```
feishu[main]: received message from ou_... in oc_ceaacf99d602be4842fe4505a77a4e70 (p2p)
feishu[main]: dispatching to agent (session=agent:main:main)
```
可以... | open | null | false | 3 | [] | [] | 2026-03-22T12:27:02Z | 2026-05-04T21:00:13Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | LiXingxun | 40,430,737 | MDQ6VXNlcjQwNDMwNzM3 | User | false |
openclaw/openclaw | 4,115,996,478 | I_kwDOQb6kR871VR8- | 52,254 | https://github.com/openclaw/openclaw/issues/52254 | https://api.github.com/repos/openclaw/openclaw/issues/52254 | [Control UI] 部分模块中文翻译缺失 | ### Summary
## 问题描述
当在 Control UI 中切换语言为中文 (zh-CN) 时，部分 Tab 内部的 UI 仍然是英文。
## 复现步骤
1. 打开 Control UI (http://<host>:18789/)
2. 在 Access card 中选择语言为 "简体中文"
3. 切换到以下 Tab，观察内部 UI 仍为英文：
- 代理 (Agents)
- 技能 (Skills)
- 频道 (Channels)
- 会话 (Sessions)
- 节点 (Nodes)
- 调试 (Debug)
## 预期行为
所有 Tab 内部 UI 都应该显示中文。
... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-22T13:01:08Z | 2026-05-04T19:11:51Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | RuCu6 | 49,463,634 | MDQ6VXNlcjQ5NDYzNjM0 | User | false |
openclaw/openclaw | 4,116,083,756 | I_kwDOQb6kR871VnQs | 52,266 | https://github.com/openclaw/openclaw/issues/52266 | https://api.github.com/repos/openclaw/openclaw/issues/52266 | [Feature Request] Add multi-session UI support for WebChat (session switching sidebar) | ### Summary
Currently, WebChat (the built-in Control UI) only supports a single active session at a time. Users cannot create, view, or switch between multiple isolated sessions (e.g., "Work", "Writing", "AI News") within the WebChat interface.
This limits productivity for users who want to:
- Separate contexts (work... | closed | not_planned | true | 3 | [
"enhancement"
] | [] | 2026-03-22T13:36:36Z | 2026-05-04T04:37:25Z | 2026-05-01T05:27:31Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | SheenChow | 17,705,905 | MDQ6VXNlcjE3NzA1OTA1 | User | false |
openclaw/openclaw | 4,116,186,321 | I_kwDOQb6kR871WATR | 52,288 | https://github.com/openclaw/openclaw/issues/52288 | https://api.github.com/repos/openclaw/openclaw/issues/52288 | Bug: only last client tool call returned when agent calls multiple tools per turn | ## Bug Description
When an Agent calls multiple client function tools in a single turn via `/v1/responses`, only the **last** tool call is returned to the caller. All preceding tool calls are silently dropped.
## Reproduction
Send a request with 3 client tools and instructions to call all 3:
```bash
curl -X POST ht... | closed | completed | true | 3 | [] | [] | 2026-03-22T14:23:34Z | 2026-05-05T04:25:58Z | 2026-05-02T23:47:47Z | CONTRIBUTOR | null | 20260505T060034Z | 2026-05-05T06:00:34Z | CharZhou | 17,255,546 | MDQ6VXNlcjE3MjU1NTQ2 | User | false |
openclaw/openclaw | 4,116,305,154 | I_kwDOQb6kR871WdUC | 52,313 | https://github.com/openclaw/openclaw/issues/52313 | https://api.github.com/repos/openclaw/openclaw/issues/52313 | gateway restart/update can fail to come back when respawn reuses unstable package-manager paths | ## Summary
Message-triggered `restart` / `update.run` can occasionally fail to bring the gateway back after shutdown.
## What is happening
The gateway run loop already tries to do a full fresh-process restart after `SIGUSR1`, but `restartGatewayProcessWithFreshPid()` currently respawns the child with `process.execArgv... | open | null | false | 2 | [] | [] | 2026-03-22T15:15:52Z | 2026-05-04T21:00:24Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | RichardCao | 4,612,401 | MDQ6VXNlcjQ2MTI0MDE= | User | false |
openclaw/openclaw | 4,116,415,966 | I_kwDOQb6kR871W4Xe | 52,353 | https://github.com/openclaw/openclaw/issues/52353 | https://api.github.com/repos/openclaw/openclaw/issues/52353 | Feature Request: Emit Agent Event on Embedded Run Timeout | # Feature Request: Emit Agent Event on Embedded Run Timeout
## Problem
When an embedded run times out, OpenClaw only logs a warning but does not emit an agent event. This makes it impossible for plugins (like task-monitor) to detect and handle timeout scenarios.
**Current behavior** (in `thread-bindings-SYAnWHuW.js`... | closed | completed | true | 2 | [] | [] | 2026-03-22T16:08:06Z | 2026-05-03T04:34:15Z | 2026-05-01T02:24:23Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | bjyounger | 263,584,108 | U_kgDOD7X5bA | User | false |
openclaw/openclaw | 4,116,502,534 | I_kwDOQb6kR871XNgG | 52,373 | https://github.com/openclaw/openclaw/issues/52373 | https://api.github.com/repos/openclaw/openclaw/issues/52373 | [Feature]:Add configUrl and configButton support in SKILL.md frontmatter | ### Summary
Add support for `configUrl` and `configButton` fields in SKILL.md frontmatter to allow skills to expose a configuration UI entry point in the OpenClaw Web Dashboard.
### Problem to solve
Currently, when a skill has a web-based configuration interface, users have no direct way to access it from the OpenCl... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-22T16:48:31Z | 2026-05-04T19:31:10Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | chenjiayang2019 | 59,195,642 | MDQ6VXNlcjU5MTk1NjQy | User | false |
openclaw/openclaw | 4,116,519,626 | I_kwDOQb6kR871XRrK | 52,382 | https://github.com/openclaw/openclaw/issues/52382 | https://api.github.com/repos/openclaw/openclaw/issues/52382 | feat: raise maxPingPongTurns ceiling above 5 or add per-workflow override | ## Summary
The platform schema currently enforces a hard ceiling of 5 on `session.agentToAgent.maxPingPongTurns`. Multi-agent workflows that involve more than 2-3 agents in a single coordination chain hit this ceiling silently — outbound replies drop with no error surfaced to the calling agent.
## Observed Behavior
... | open | null | false | 4 | [] | [] | 2026-03-22T16:56:04Z | 2026-05-04T21:00:28Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,116,520,424 | I_kwDOQb6kR871XR3o | 52,384 | https://github.com/openclaw/openclaw/issues/52384 | https://api.github.com/repos/openclaw/openclaw/issues/52384 | feat: circuit breaker on auto-resume injection during consecutive model errors | ## Summary
When a session experiences consecutive model errors (e.g., during a 429 flood), the platform's automatic "Continue where you left off" resume prompt continues injecting into the failing session. This amplifies token burn, adds noise to the session history, and does not help recovery.
## Observed Behavior
... | closed | not_planned | true | 3 | [] | [] | 2026-03-22T16:56:25Z | 2026-05-03T04:34:15Z | 2026-05-01T02:24:29Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,116,520,831 | I_kwDOQb6kR871XR9_ | 52,385 | https://github.com/openclaw/openclaw/issues/52385 | https://api.github.com/repos/openclaw/openclaw/issues/52385 | feat: sessions_send inter-agent messages should not render in user-facing Slack threads | ## Summary
When `sessions_send` is used between agents (agent-to-agent coordination), the message appears as a visible reply in the human-facing Slack thread. This pollutes the user's view with internal coordination traffic and breaks the UI for any thread the user is actively watching.
## Observed Behavior
On 2026-... | open | null | false | 3 | [] | [] | 2026-03-22T16:56:36Z | 2026-05-04T20:16:12Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,116,577,020 | I_kwDOQb6kR871Xfr8 | 52,396 | https://github.com/openclaw/openclaw/issues/52396 | https://api.github.com/repos/openclaw/openclaw/issues/52396 | Discord retry policy only covers 429 — misses 502/503/timeout/connection errors | ## Problem
Discord's outbound message retry (`extensions/discord/src/retry.ts`) only retries on `RateLimitError` (HTTP 429). Any other transient failure — 502, 503, connection reset, timeout, DNS resolution failure — is not retried and fails immediately.
Compare to Telegram's retry policy (`src/infra/retry-policy.ts`... | closed | completed | true | 4 | [] | [] | 2026-03-22T17:19:17Z | 2026-05-04T04:34:43Z | 2026-05-02T01:47:23Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | sunshineo | 1,072,766 | MDQ6VXNlcjEwNzI3NjY= | User | false |
openclaw/openclaw | 4,116,689,333 | I_kwDOQb6kR871X7G1 | 52,412 | https://github.com/openclaw/openclaw/issues/52412 | https://api.github.com/repos/openclaw/openclaw/issues/52412 | [Feature]: Plugin API for aborting agent runs (api.runtime.agent.abort) | ### Summary
Add `api.runtime.agent.abort({ runId })` to the plugin runtime, allowing plugins to cancel in-progress agent runs by runId.
### Problem to solve
There is currently no way for plugins to programmatically cancel an in-progress agent run. Once an agent starts processing, it runs to completion regardless of ... | open | null | false | 1 | [] | [] | 2026-03-22T18:09:57Z | 2026-05-04T19:31:17Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | livingghost | 14,040,213 | MDQ6VXNlcjE0MDQwMjEz | User | false |
openclaw/openclaw | 4,116,732,123 | I_kwDOQb6kR871YFjb | 52,421 | https://github.com/openclaw/openclaw/issues/52421 | https://api.github.com/repos/openclaw/openclaw/issues/52421 | LLM API error: unexpected tool_use_id in tool_result blocks causes session disruption | ## Summary
Intermittent but recurring API error that disrupts active sessions:
```
LLM request rejected: messages.6.content.0: unexpected `tool_use_id` found in `tool_result` blocks: toolu01U1J4aPKsdDDM2CwDXP9Red. Each `tool_result` block must have a corresponding `tool_use` block in the previous message.
```
## Envi... | open | null | false | 6 | [] | [] | 2026-03-22T18:29:33Z | 2026-05-04T19:31:18Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | joesinvestments | 257,641,472 | U_kgDOD1tMAA | User | false |
openclaw/openclaw | 4,116,852,446 | I_kwDOQb6kR871Yi7e | 52,442 | https://github.com/openclaw/openclaw/issues/52442 | https://api.github.com/repos/openclaw/openclaw/issues/52442 | WhatsApp listener close() calls sock.ws?.close() instead of sock.end() — causes zombie sockets | ## Bug Description
The WhatsApp listener's `close()` function in `extensions/whatsapp/src/inbound/monitor.ts:481` calls `sock.ws?.close()` instead of `sock.end()`. This bypasses Baileys' internal cleanup, causing zombie sockets that report "connected" but cannot receive inbound messages.
## Root Cause
`sock.end()` (... | closed | completed | true | 5 | [] | [] | 2026-03-22T19:09:10Z | 2026-05-04T04:34:46Z | 2026-05-02T01:29:54Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | essendigitalgroup-cyber | 254,002,603 | U_kgDODyPFqw | User | false |
openclaw/openclaw | 4,346,293,463 | I_kwDOQb6kR88AAAABAw8s1w | 73,804 | https://github.com/openclaw/openclaw/issues/73804 | https://api.github.com/repos/openclaw/openclaw/issues/73804 | Bonjour mDNS probe causes supervisor double-boot on macOS (+30s startup) | ## Bug Description
On macOS (Mac Mini M4 Pro, macOS 15.x), the bundled Bonjour/mDNS plugin triggers a probe/announce loop that causes the gateway supervisor to restart the process a second time, adding ~30 seconds to every gateway startup.
## Steps to Reproduce
1. Run OpenClaw gateway via launchd (or any process sup... | closed | completed | true | 2 | [] | [] | 2026-04-28T21:19:24Z | 2026-05-03T04:33:43Z | 2026-04-30T14:59:19Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | dertbv | 3,845,627 | MDQ6VXNlcjM4NDU2Mjc= | User | false |
openclaw/openclaw | 4,117,068,459 | I_kwDOQb6kR871ZXqr | 52,473 | https://github.com/openclaw/openclaw/issues/52473 | https://api.github.com/repos/openclaw/openclaw/issues/52473 | [Feature]: Auto-inject SKILL.md content when skill is invoked via slash command | ## Problem
When a user invokes a skill via `/skill-name` (slash command), OpenClaw forwards the message to the model as a normal request. The model must then **voluntarily** call `read` on the SKILL.md file to load the skill instructions. There is no mechanism that forces injection of the skill content into context.
... | open | null | false | 1 | [] | [] | 2026-03-22T20:44:50Z | 2026-05-04T19:31:25Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | lysanderparodi-beep | 260,368,818 | U_kgDOD4Tpsg | User | false |
openclaw/openclaw | 4,117,274,546 | I_kwDOQb6kR871aJ-y | 52,496 | https://github.com/openclaw/openclaw/issues/52496 | https://api.github.com/repos/openclaw/openclaw/issues/52496 | feat: sessions_send should refuse (or warn) when target is a thread session key | ## Summary
`sessions_send` currently delivers inter-agent messages to thread session keys (keys containing `:thread:`) without error or warning. This causes inter-agent coordination traffic to appear as `role: user` messages inside an active human-facing conversation, breaking the user's Slack thread UI.
## Incident
... | closed | completed | true | 3 | [] | [] | 2026-03-22T22:30:07Z | 2026-05-05T04:28:50Z | 2026-05-02T04:56:41Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,117,278,066 | I_kwDOQb6kR871aK1y | 52,498 | https://github.com/openclaw/openclaw/issues/52498 | https://api.github.com/repos/openclaw/openclaw/issues/52498 | [Feature]: Show who session_send | ### Summary
In sessions of agents, I can see that agents do communicate with each other, but on message level I do not see which agent sent the message.
### Problem to solve
If you have a lot of agents talking to each other, it's handy to be able to see (and maybe filter) who sent what to a particular session.
###... | open | null | false | 3 | [
"enhancement"
] | [] | 2026-03-22T22:32:06Z | 2026-05-04T20:33:48Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | najef1979-code | 233,555,716 | U_kgDODevHBA | User | false |
openclaw/openclaw | 4,117,278,860 | I_kwDOQb6kR871aLCM | 52,500 | https://github.com/openclaw/openclaw/issues/52500 | https://api.github.com/repos/openclaw/openclaw/issues/52500 | fix: config schema ceiling rejections should explicitly state the maximum allowed value | ## Summary
When a config value is rejected because it exceeds a schema-enforced ceiling, the error message does not state what the maximum allowed value is. The operator is told their value was rejected but not what value would be accepted.
## Observed Behavior
Attempting to set `session.agentToAgent.maxPingPongTurn... | open | null | false | 3 | [] | [] | 2026-03-22T22:32:29Z | 2026-05-04T19:31:33Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,117,416,435 | I_kwDOQb6kR871asnz | 52,526 | https://github.com/openclaw/openclaw/issues/52526 | https://api.github.com/repos/openclaw/openclaw/issues/52526 | [Bug]: agent --json returns pre-hook assistant text instead of finalized before_message_write output | ## Summary
On OpenClaw `2026.3.22` (`3286de3`), `node dist/index.js agent --json` can return the model's raw final text instead of the finalized assistant message after plugin `before_message_write` hooks run.
The persisted session transcript contains the hook-mutated assistant message, but the CLI JSON result surfac... | open | null | false | 3 | [] | [] | 2026-03-22T23:51:03Z | 2026-05-04T21:00:35Z | null | CONTRIBUTOR | null | 20260505T000034Z | 2026-05-05T00:00:34Z | ninjaa | 1,315,093 | MDQ6VXNlcjEzMTUwOTM= | User | false |
openclaw/openclaw | 4,117,518,662 | I_kwDOQb6kR871bFlG | 52,540 | https://github.com/openclaw/openclaw/issues/52540 | https://api.github.com/repos/openclaw/openclaw/issues/52540 | [Bug]: 1password skill instructs tmux-only usage, but tmux breaks desktop app integration on macOS | ### Bug type
Bundled skill gives incorrect guidance
### Summary
The bundled `1password` skill (`skills/1password/SKILL.md`) states:
> **REQUIRED: create a fresh tmux session for all `op` commands (no direct `op` calls outside tmux).**
> **Do not run `op` outside tmux; stop and ask if tmux is unavailable.**
This is... | open | null | false | 2 | [] | [] | 2026-03-23T00:39:18Z | 2026-05-04T19:31:43Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | tylerbittner | 22,528,533 | MDQ6VXNlcjIyNTI4NTMz | User | false |
openclaw/openclaw | 4,117,688,823 | I_kwDOQb6kR871bvH3 | 52,568 | https://github.com/openclaw/openclaw/issues/52568 | https://api.github.com/repos/openclaw/openclaw/issues/52568 | [Feature]: Discord thread-reply mention gate — respond to thread starter freely, require mention for follow-up replies | ## Summary
Add a per-channel config option that controls mention requirements differently for thread-starter messages vs follow-up replies within a thread.
## Current Behavior
With `requireMention: false` + `autoThread: true`: agent responds to all messages including every follow-up reply in threads.
With `requireM... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-23T01:49:22Z | 2026-05-04T19:31:45Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | jasonhuhx | 1,328,495 | MDQ6VXNlcjEzMjg0OTU= | User | false |
openclaw/openclaw | 4,117,727,594 | I_kwDOQb6kR871b4lq | 52,573 | https://github.com/openclaw/openclaw/issues/52573 | https://api.github.com/repos/openclaw/openclaw/issues/52573 | [Bug]: Kimi web_search often degrades into generic chat answers instead of actual search results | ## Summary
When `tools.web.search.provider` is set to `kimi`, `web_search` often degrades into an AI-style generic answer instead of returning actual search results. In practice it can say things like “I can't predict the future” or “I don't have browsing capability” even though the provider is configured as a web sear... | closed | completed | true | 4 | [] | [] | 2026-03-23T02:06:43Z | 2026-05-05T04:28:37Z | 2026-05-02T06:03:54Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | wangwllu | 7,668,944 | MDQ6VXNlcjc2Njg5NDQ= | User | false |
openclaw/openclaw | 4,117,748,031 | I_kwDOQb6kR871b9k_ | 52,577 | https://github.com/openclaw/openclaw/issues/52577 | https://api.github.com/repos/openclaw/openclaw/issues/52577 | Queue inbound Discord messages during WS reconnect window (~22s post-boot gap) | ## Summary
After a gateway restart, there's a ~22-second window where the HTTP server and cron scheduler are running but the Discord WebSocket isn't connected yet. Messages sent by users during this window are silently dropped — no error, no retry, no indication to the user that the bot didn't receive their message.
... | open | null | false | 1 | [] | [] | 2026-03-23T02:14:43Z | 2026-05-04T19:31:49Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Stache73 | 7,196,384 | MDQ6VXNlcjcxOTYzODQ= | User | false |
openclaw/openclaw | 4,117,886,819 | I_kwDOQb6kR871cfdj | 52,596 | https://github.com/openclaw/openclaw/issues/52596 | https://api.github.com/repos/openclaw/openclaw/issues/52596 | Feature: Session authentication PIN gate (framework-level tool blocking) | ## Problem
Currently, session authentication (e.g. PIN verification) is purely LLM-enforced via workspace context files (SOUL.md, AGENTS.md). This provides no real security against prompt injection attacks from external content (websites, PDFs, malicious files) that could instruct the agent to skip authentication and ... | open | null | false | 2 | [] | [] | 2026-03-23T03:08:26Z | 2026-05-04T19:31:50Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | avaloki108 | 111,966,879 | U_kgDOBqx6nw | User | false |
openclaw/openclaw | 4,118,080,082 | I_kwDOQb6kR871dOpS | 52,618 | https://github.com/openclaw/openclaw/issues/52618 | https://api.github.com/repos/openclaw/openclaw/issues/52618 | fix(feishu): WebSocket connection not recovered after network disruption | ## Bug Description
When the Feishu WebSocket connection drops due to a network disruption, the connection is never re-established. The gateway silently loses the ability to receive Feishu messages until a manual `openclaw gateway restart`.
## Steps to Reproduce
1. Start OpenClaw with Feishu channel enabled (`connect... | closed | completed | true | 2 | [] | [] | 2026-03-23T04:05:09Z | 2026-05-04T04:36:20Z | 2026-05-01T13:27:28Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | bridzl | 52,784,198 | MDQ6VXNlcjUyNzg0MTk4 | User | false |
openclaw/openclaw | 4,118,385,867 | I_kwDOQb6kR871eZTL | 52,648 | https://github.com/openclaw/openclaw/issues/52648 | https://api.github.com/repos/openclaw/openclaw/issues/52648 | google-vertex provider authentication fails when running through proxychains | ## Bug Description
OpenClaw's `google-vertex` provider consistently fails with `401 CREDENTIALS_MISSING` authentication errors when the gateway service runs through `proxychains`.
**Key findings:**
1. ✅ Service Account credentials are valid and work directly
2. ✅ `gcloud` ADC authentication succeeds without proxy
3.... | open | null | false | 1 | [] | [] | 2026-03-23T05:37:23Z | 2026-05-04T21:00:39Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | seb3mz | 204,283,073 | U_kgDODC0cwQ | User | false |
openclaw/openclaw | 4,118,547,648 | I_kwDOQb6kR871fAzA | 52,662 | https://github.com/openclaw/openclaw/issues/52662 | https://api.github.com/repos/openclaw/openclaw/issues/52662 | [Feature]: Browser sandbox should support non-Docker backends (external CDP/noVNC endpoint) | ### Summary
Allow the browser sandbox to connect to an externally managed browser container via configured CDP/noVNC endpoints, bypassing Docker daemon dependency.
### Problem to solve
The sandbox backend abstraction (`SandboxBackendHandle`) supports Docker, SSH, and OpenShell backends, allowing the gateway to run w... | open | null | false | 3 | [] | [] | 2026-03-23T06:20:07Z | 2026-05-04T19:32:01Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | carrotRakko | 24,588,751 | MDQ6VXNlcjI0NTg4NzUx | User | false |
openclaw/openclaw | 4,118,598,343 | I_kwDOQb6kR871fNLH | 52,665 | https://github.com/openclaw/openclaw/issues/52665 | https://api.github.com/repos/openclaw/openclaw/issues/52665 | [Feature]: Optimization for Dockerfile.sandbox-common # add retries and fix-missings for apt-get | ### Summary
# add retries and fix-missings for apt-get to make oepnclaw-sandbox-cache
### Problem to solve
Run sandbox-common-set to build openclaw image but easily failed because the occasionally bad network.
### Proposed solution
Support upmost 5 retries to get dependencies.
Support to use fix-missings to get de... | open | reopened | false | 1 | [
"enhancement"
] | [] | 2026-03-23T06:32:28Z | 2026-05-04T19:32:05Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | broorr | 66,622,308 | MDQ6VXNlcjY2NjIyMzA4 | User | false |
openclaw/openclaw | 4,119,131,864 | I_kwDOQb6kR871hPbY | 52,732 | https://github.com/openclaw/openclaw/issues/52732 | https://api.github.com/repos/openclaw/openclaw/issues/52732 | Per-agent compaction and contextPruning overrides in agents.list | ## Problem
Currently `compaction` and `contextPruning` settings can only be configured globally in `agents.defaults`. When using `agents.list` with per-agent overrides (e.g., `model`, `workspace`), these two settings are rejected as unrecognized keys:
```
agents.list.0: Unrecognized keys: "compaction", "contextPrunin... | open | null | false | 1 | [] | [] | 2026-03-23T08:25:55Z | 2026-05-04T21:00:41Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | adrijc93 | 213,996,107 | U_kgDODMFSSw | User | false |
openclaw/openclaw | 4,119,184,917 | I_kwDOQb6kR871hcYV | 52,736 | https://github.com/openclaw/openclaw/issues/52736 | https://api.github.com/repos/openclaw/openclaw/issues/52736 | [Feature]: Add screenshot capture button to Control UI chat input (like Doubao) | ### Summary
Add a screenshot capture button to the Control UI chat input area, similar to Doubao (豆包) app. Users should be able to:
1. Click a scissors/screenshot icon button
2. Capture a region of the screen
3. See a preview thumbnail in the input area
4. Send the screenshot to the AI for analysis
### Problem to sol... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-23T08:35:16Z | 2026-05-04T19:32:08Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | taot3661-spec | 239,262,530 | U_kgDODkLbQg | User | false |
openclaw/openclaw | 4,119,463,465 | I_kwDOQb6kR871igYp | 52,757 | https://github.com/openclaw/openclaw/issues/52757 | https://api.github.com/repos/openclaw/openclaw/issues/52757 | GPT-5.4 default parameter values break message tool (poll, components, media path) | ## Summary
When using GPT-5.4 as the model provider, the message tool fails in multiple ways because GPT-5.4 sends **all** tool parameters with default values (e.g. `pollDurationHours: 24`, `pollMulti: false`, `components: {blocks:[], modal:{fields:[],...}, ...}`), unlike other models that omit unused optional paramet... | open | null | false | 2 | [] | [] | 2026-03-23T09:23:11Z | 2026-05-04T19:32:12Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | skernelx | 250,769,487 | U_kgDODvJwTw | User | false |
openclaw/openclaw | 4,119,947,633 | I_kwDOQb6kR871kWlx | 52,789 | https://github.com/openclaw/openclaw/issues/52789 | https://api.github.com/repos/openclaw/openclaw/issues/52789 | Feature: Add delivery.mode="session" for cron jobs (internal agent-to-agent delivery via sessions_send) | ## Summary
Cron jobs currently support three delivery modes: `announce` (posts to a chat channel), `webhook` (HTTP POST), and `none` (agent handles delivery itself via tools).
There is no built-in delivery mode that routes cron job output to another session internally via `sessions_send`. This creates a gap for workf... | open | null | false | 1 | [] | [] | 2026-03-23T10:48:43Z | 2026-05-04T21:00:45Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | rm289 | 961,275 | MDQ6VXNlcjk2MTI3NQ== | User | false |
openclaw/openclaw | 4,120,026,794 | I_kwDOQb6kR871kp6q | 52,793 | https://github.com/openclaw/openclaw/issues/52793 | https://api.github.com/repos/openclaw/openclaw/issues/52793 | [Feature]: Granular `tool.fs.workspaceOnly` configuration per tool | ### Summary
`tool.fs.workspaceOnly` is currently a single global boolean for filesystem tools.
When set to `true`, it restricts **all** fs-related tools (including read operations) to the workspace root.
In real-world setups, this is too coarse-grained. We often need:
- strict restrictions for write/edit/patch oper... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-23T11:03:22Z | 2026-05-04T21:00:47Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | Noone0013 | 110,889,386 | U_kgDOBpwJqg | User | false |
openclaw/openclaw | 4,120,148,034 | I_kwDOQb6kR871lHhC | 52,802 | https://github.com/openclaw/openclaw/issues/52802 | https://api.github.com/repos/openclaw/openclaw/issues/52802 | openai-node streaming: event-only SSE flush causes JSON.parse('') crash (Unexpected end of JSON input) | ### Summary
When consuming an SSE stream, if an SSE “message” is flushed with an `event:` field but **no `data:` lines** (i.e. `event: …` followed by a blank line), the OpenAI Node SDK attempts `JSON.parse` on an empty string and throws:
- `Unexpected end of JSON input`
This shows up downstream in OpenClaw as intermi... | closed | completed | true | 2 | [] | [] | 2026-03-23T11:24:20Z | 2026-05-01T04:40:52Z | 2026-04-29T02:29:04Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | LyHug | 40,082,700 | MDQ6VXNlcjQwMDgyNzAw | User | false |
openclaw/openclaw | 4,120,154,340 | I_kwDOQb6kR871lJDk | 52,803 | https://github.com/openclaw/openclaw/issues/52803 | https://api.github.com/repos/openclaw/openclaw/issues/52803 | [Feature]: Improve Control UI for Multi-Agent + Subagent Orchestration (Hierarchy, Bulk Ops, and Scalability) | ### Summary
Improve Control UI to support scalable multi-agent orchestration with clear agent-subagent hierarchy, active-first visibility, and efficient bulk operations.
### Problem to solve
When multiple agents and subagents run concurrently, Control UI becomes hard to operate at scale.
Current views are too flat,... | open | null | false | 4 | [
"enhancement"
] | [] | 2026-03-23T11:25:34Z | 2026-05-01T21:59:24Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | Leeon-K | 48,170,673 | MDQ6VXNlcjQ4MTcwNjcz | User | false |
openclaw/openclaw | 4,120,400,257 | I_kwDOQb6kR871mFGB | 52,826 | https://github.com/openclaw/openclaw/issues/52826 | https://api.github.com/repos/openclaw/openclaw/issues/52826 | Add structured process events for sessions | ClawWork needs structured process events from upstream.
Current chat/tool/lifecycle events are not enough for a real process system. The client should not derive progress from assistant text.
Please add session-level process events with stable step identity.
Required:
- process.started
- process.replanned
- step.cr... | open | null | false | 1 | [] | [] | 2026-03-23T12:13:42Z | 2026-05-04T19:32:29Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | samzong | 13,782,141 | MDQ6VXNlcjEzNzgyMTQx | User | false |
openclaw/openclaw | 4,120,640,784 | I_kwDOQb6kR871m_0Q | 52,849 | https://github.com/openclaw/openclaw/issues/52849 | https://api.github.com/repos/openclaw/openclaw/issues/52849 | [Feature]: Add before_tool_result_emit hook for live tool-result mutation before same-turn model consumption | ### Summary
OpenClaw currently exposes strong hook surfaces around tool execution, but there appears to be a missing seam for security/compliance use cases that need
to inspect or transform a tool result before it is consumed by the model in the same turn.
Today, plugins/runtime seem to support:
- before_tool_cal... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-03-23T12:51:29Z | 2026-05-04T19:32:30Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | sheldonade76 | 267,317,764 | U_kgDOD-7yBA | User | false |
openclaw/openclaw | 4,120,735,632 | I_kwDOQb6kR871nW-Q | 52,859 | https://github.com/openclaw/openclaw/issues/52859 | https://api.github.com/repos/openclaw/openclaw/issues/52859 | Telegram bold/italic entities not rendered as markdown in agent prompt (regression) | ## Bug Report
**Version:** OpenClaw 2026.3.13 (61d171a)
**Channel:** Telegram (bot DMs)
### Description
Telegram message formatting (bold, italic, code, strikethrough) entities are not being converted to markdown in the text passed to the agent prompt. Only `text_link` entities are processed via `expandTextLinks()`... | open | null | false | 2 | [] | [] | 2026-03-23T13:05:58Z | 2026-05-04T19:32:32Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | egida307-star | 269,931,985 | U_kgDOEBbV0Q | User | false |
openclaw/openclaw | 4,121,634,159 | I_kwDOQb6kR871qyVv | 52,960 | https://github.com/openclaw/openclaw/issues/52960 | https://api.github.com/repos/openclaw/openclaw/issues/52960 | [Feature]: Control UI should show compact run progress for long or multi-agent tasks | ## Summary
The Control UI needs a compact run-progress indicator for long or multi-agent tasks. Right now a run can appear stalled even when OpenClaw is still making progress or has already finished internal work but has not yet surfaced the final reply.
This is not a request for verbose logs. The goal is a low-noise... | open | null | false | 2 | [] | [] | 2026-03-23T15:14:18Z | 2026-05-04T19:40:50Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | betamod | 8,882,726 | MDQ6VXNlcjg4ODI3MjY= | User | false |
openclaw/openclaw | 4,122,037,575 | I_kwDOQb6kR871sU1H | 52,997 | https://github.com/openclaw/openclaw/issues/52997 | https://api.github.com/repos/openclaw/openclaw/issues/52997 | Forum groups: option for shared main session with topic metadata | ## Problem
In Telegram forum groups (supergroups with topics enabled), each forum topic currently creates an isolated session (`:topic:<threadId>` appended to session key). Similarly, DM threads use `message_thread_id` to create separate sessions per thread. This is documented and intentional.
However, for power user... | open | null | false | 1 | [] | [] | 2026-03-23T16:14:03Z | 2026-05-04T19:32:45Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | thealfredbot | 258,442,359 | U_kgDOD2eEdw | User | false |
openclaw/openclaw | 4,139,168,949 | I_kwDOQb6kR872trS1 | 54,749 | https://github.com/openclaw/openclaw/issues/54749 | https://api.github.com/repos/openclaw/openclaw/issues/54749 | Mobile: Clicking report in list shows Command Overview instead of report content | ## Bug Description
On mobile devices, when navigating to the Reports section and clicking on a report from the list, the UI incorrectly displays the Command Overview instead of the actual report content.
## Steps to Reproduce
1. Open OpenClaw Control UI on a mobile device (or mobile viewport)
2. Navigate to Reports se... | open | null | false | 1 | [] | [] | 2026-03-25T23:07:39Z | 2026-05-04T21:42:05Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | fortifyos | 263,220,758 | U_kgDOD7BuFg | User | false |
openclaw/openclaw | 4,139,644,437 | I_kwDOQb6kR872vfYV | 54,797 | https://github.com/openclaw/openclaw/issues/54797 | https://api.github.com/repos/openclaw/openclaw/issues/54797 | Control UI: Add 'clear completed' for subagent sessions in session dropdown | ## Problem
The session dropdown in the Control UI mixes active sessions (main chat, Telegram channels, heartbeat) with completed background subagent runs. Once a subagent finishes, its entry remains in the dropdown until the session maintenance prune window expires (which could be hours or days).
There is currently n... | open | null | false | 1 | [] | [] | 2026-03-26T01:15:43Z | 2026-05-04T21:34:17Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | aakashbadiani | 39,374,949 | MDQ6VXNlcjM5Mzc0OTQ5 | User | false |
openclaw/openclaw | 4,139,644,545 | I_kwDOQb6kR872vfaB | 54,798 | https://github.com/openclaw/openclaw/issues/54798 | https://api.github.com/repos/openclaw/openclaw/issues/54798 | feat(cron): shared atomic counter for concurrent cron sub-agents (daily rate limits) | ## Problem
When multiple cron jobs spawn concurrent sub-agents that each need to read-modify-write a shared counter (e.g., daily post/review counts), there is no built-in atomic primitive. Each agent reads state, modifies, and writes — creating a race window.
**Observed symptom:** Two concurrent engage cycles both re... | open | null | false | 1 | [] | [] | 2026-03-26T01:15:44Z | 2026-05-04T21:42:20Z | null | NONE | null | 20260505T000034Z | 2026-05-05T00:00:34Z | gasvn | 20,515,144 | MDQ6VXNlcjIwNTE1MTQ0 | User | false |
openclaw/openclaw | 4,162,666,762 | I_kwDOQb6kR874HUEK | 56,815 | https://github.com/openclaw/openclaw/issues/56815 | https://api.github.com/repos/openclaw/openclaw/issues/56815 | memorySearch: embedding reindex fails with 'TypeError: fetch failed' after indexing ~40K chunks | ## Description
Memory search embedding reindex consistently fails with `TypeError: fetch failed` after successfully indexing a significant number of chunks (~41K out of estimated ~45K). The `.tmp` file is deleted on failure (runSafeReindex rollback), so all progress is lost and the next attempt starts from scratch — c... | open | null | false | 1 | [] | [] | 2026-03-29T05:14:08Z | 2026-05-01T03:52:44Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dionysoslin615 | 214,426,612 | U_kgDODMfj9A | User | false |
openclaw/openclaw | 4,164,089,856 | I_kwDOQb6kR874MvgA | 57,017 | https://github.com/openclaw/openclaw/issues/57017 | https://api.github.com/repos/openclaw/openclaw/issues/57017 | fix: readCronRunLogEntriesPage reads entire JSONL into memory without size cap | ## Summary
`readCronRunLogEntriesPage` reads the entire JSONL log file into memory with `fs.readFile` before parsing and paginating. While writes are pruned to `DEFAULT_CRON_RUN_LOG_MAX_BYTES` (2MB), the pruning is async and can fail silently under disk pressure. If pruning fails, the file grows unbounded, and a subse... | open | null | false | 2 | [] | [] | 2026-03-29T12:32:00Z | 2026-05-01T06:04:23Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jamie8johnson | 85,816,907 | MDQ6VXNlcjg1ODE2OTA3 | User | false |
openclaw/openclaw | 4,165,072,021 | I_kwDOQb6kR874QfSV | 57,169 | https://github.com/openclaw/openclaw/issues/57169 | https://api.github.com/repos/openclaw/openclaw/issues/57169 | [Bug]: macOS node advertises screen capability but runtime blocks screen.record via platform allowlist | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
On OpenClaw 2026.3.28, a macOS node can be paired, connected, granted Screen Recording permission, and report screen capability, but screen.record is still rejected at runtime with:
node command not allowed: "... | open | null | false | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-29T16:49:36Z | 2026-05-01T03:52:45Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | JoseloCaropreso | 164,774,475 | U_kgDOCdJCSw | User | false |
openclaw/openclaw | 4,168,826,397 | I_kwDOQb6kR874ez4d | 57,594 | https://github.com/openclaw/openclaw/issues/57594 | https://api.github.com/repos/openclaw/openclaw/issues/57594 | Zalo channel rejects valid non-numeric chat_id values for proactive outbound sends | ## Summary
`openclaw message send --channel zalo --target <id>` rejects valid Zalo Bot Platform `chat_id` values that are non-numeric (hex-like), even though upstream Zalo Bot API accepts them and delivers successfully.
This causes a mismatch where:
- replying inside an existing Zalo session works
- proactive outboun... | open | null | false | 1 | [] | [] | 2026-03-30T08:48:39Z | 2026-05-01T06:04:29Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | malayvuong | 13,240,975 | MDQ6VXNlcjEzMjQwOTc1 | User | false |
openclaw/openclaw | 4,204,612,186 | I_kwDOQb6kR876nUpa | 60,826 | https://github.com/openclaw/openclaw/issues/60826 | https://api.github.com/repos/openclaw/openclaw/issues/60826 | config-audit.jsonl logs plaintext secrets in CLI argv | ## Description
The config tamper detection system (`config-audit.jsonl`) records full CLI argument vectors when config changes are detected. These argv entries contain plaintext secrets (gateway tokens, bot tokens, API keys) that were passed as command-line arguments or present in the process environment.
A security ... | closed | completed | true | 1 | [] | [] | 2026-04-04T11:08:43Z | 2026-05-03T04:34:18Z | 2026-05-01T02:08:16Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | Belthezar46 | 85,139,005 | MDQ6VXNlcjg1MTM5MDA1 | User | false |
openclaw/openclaw | 4,221,751,513 | I_kwDOQb6kR877otDZ | 62,863 | https://github.com/openclaw/openclaw/issues/62863 | https://api.github.com/repos/openclaw/openclaw/issues/62863 | edit 工具稳定性改进：4 重防护机制 | ## 问题背景
edit 工具要求 oldText 参数与文件内容**逐字符精确匹配**（包括空格、换行），稍有差异就失败。实际使用中经常因 whitespace 不一致导致编辑失败，触发错误报告。
## 建议方案
### 1. 空白归一化兜底（核心改进）
在匹配 oldText 之前，对「文件内容」和「oldText」都做空白归一化：
- 统一换行符（\r\n → \n）
- 多个空格/制表符压缩为 1 个
- 去掉首尾空白
- 可选：忽略空行差异
用归一化后的内容进行匹配，即使有格式差异也能成功。
### 2. 前置校验步骤
执行 edit 前自动校验：
- 用 read 读取 oldText 对应的文件内容
- 与 ... | open | null | false | 1 | [] | [] | 2026-04-08T02:51:04Z | 2026-05-02T17:54:54Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | GH66688 | 247,092,814 | U_kgDODrpWTg | User | false |
openclaw/openclaw | 4,224,714,690 | I_kwDOQb6kR8770AfC | 63,162 | https://github.com/openclaw/openclaw/issues/63162 | https://api.github.com/repos/openclaw/openclaw/issues/63162 | [Feature]: File Explorer Toggle in Control UI | ### Summary
## 🗂️ Feature Request: File Explorer Toggle in Control UI
### Description
Add a file explorer button next to the "Cron" button in the Control UI top bar. Clicking the button toggles a sidebar showing the workspace file tree.
### Use Case
- Quick file navigation without leaving the chat interface
- Brow... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-08T13:04:54Z | 2026-05-02T17:54:56Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | Ricky-bizmaker | 172,873,496 | U_kgDOCk3XGA | User | false |
openclaw/openclaw | 4,226,351,554 | I_kwDOQb6kR8776QHC | 63,272 | https://github.com/openclaw/openclaw/issues/63272 | https://api.github.com/repos/openclaw/openclaw/issues/63272 | feat(cli): `skills uninstall` design proposal + prior art reconciliation | ## Summary
Add an `openclaw skills uninstall <slug>` CLI command that cleanly removes a ClawHub-installed skill from the active workspace. Mirrors the existing `openclaw skills install` flow and the `plugins uninstall` pattern in `src/plugins/uninstall.ts`.
Scope is deliberately narrow: **workspace directory + `.claw... | open | null | false | 1 | [] | [] | 2026-04-08T17:21:44Z | 2026-05-02T18:31:13Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | EchoOfZion | 212,151,438 | U_kgDODKUsjg | User | false |
openclaw/openclaw | 4,228,736,670 | I_kwDOQb6kR878DWae | 63,471 | https://github.com/openclaw/openclaw/issues/63471 | https://api.github.com/repos/openclaw/openclaw/issues/63471 | [Feature]: Be friendlier with Dockge/etc | ### Summary
```
openclaw-gateway-1 | 2026-04-09T01:10:57.108+00:00 [gateway] loading configuration…
openclaw-gateway-1 | 2026-04-09T01:10:57.112+00:00 [gateway] resolving authentication…
openclaw-gateway-1 | 2026-04-09T01:10:57.114+00:00 Missing config. Run `openclaw setup` or set gateway.mode=local (or pass --allo... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-09T01:22:39Z | 2026-05-02T18:31:14Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | bilogic | 946,010 | MDQ6VXNlcjk0NjAxMA== | User | false |
openclaw/openclaw | 4,228,912,213 | I_kwDOQb6kR878EBRV | 63,491 | https://github.com/openclaw/openclaw/issues/63491 | https://api.github.com/repos/openclaw/openclaw/issues/63491 | [Bug]: Windows Scheduled Task gateway restart/health becomes inconsistent after ready | # [Bug]: Windows Scheduled Task gateway restart/health becomes inconsistent after ready; mixes known probe false negatives with cron/session stale state and post-ready HTTP loss
## Bug type
Regression (worked before, now fails)
## Beta release blocker
No
## Summary
On Windows with the gateway installed as a Sched... | open | null | false | 1 | [] | [] | 2026-04-09T02:15:25Z | 2026-05-02T17:31:08Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | ufomaker | 20,527,392 | MDQ6VXNlcjIwNTI3Mzky | User | false |
openclaw/openclaw | 4,110,475,913 | I_kwDOQb6kR871AOKJ | 51,312 | https://github.com/openclaw/openclaw/issues/51312 | https://api.github.com/repos/openclaw/openclaw/issues/51312 | Support per-agent userTimezone override | ## Problem
`agents.defaults.userTimezone` is global — all agents share the same timezone. In a multi-agent setup where agents serve users in different timezones (e.g. a coaching platform with coaches in US Eastern, US Pacific, and UK), there's no way to set timezone per agent.
## Current workaround
Timezone instruct... | closed | not_planned | true | 3 | [] | [] | 2026-03-20T22:51:50Z | 2026-05-01T04:41:10Z | 2026-04-28T23:37:34Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dobby1738assistant | 260,547,444 | U_kgDOD4ejdA | User | false |
openclaw/openclaw | 4,346,251,417 | I_kwDOQb6kR88AAAABAw6ImQ | 73,802 | https://github.com/openclaw/openclaw/issues/73802 | https://api.github.com/repos/openclaw/openclaw/issues/73802 | Discord exec approval card/buttons not delivered in v2026.4.26 (regression after #71804) | ## Summary
Discord exec approval requests are created and can be approved manually with `/approve ...`, but the native Discord approval card/buttons do not surface. This appears to be a regression or incomplete fix after #71804.
## Environment
- OpenClaw version: `2026.4.26`
- Channel: Discord embedded channel
- Gat... | open | null | false | 3 | [] | [] | 2026-04-28T21:11:43Z | 2026-05-03T01:23:00Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | sanchezm86 | 66,266,304 | MDQ6VXNlcjY2MjY2MzA0 | User | false |
openclaw/openclaw | 4,346,591,725 | I_kwDOQb6kR88AAAABAxO57Q | 73,836 | https://github.com/openclaw/openclaw/issues/73836 | https://api.github.com/repos/openclaw/openclaw/issues/73836 | [Bug]: Control UI/Gateway responsiveness regression during active runs: UI reconnect stalls, Telegram typing gaps, heartbeat poll noise, and media mirror UX issues | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading to v2026.4.26, OpenClaw becomes partially unresponsive while an agent run is active. The model eventually responds and Telegram delivery works, but the Control UI can fail to load/reconnect during the run, Te... | open | null | false | 1 | [
"bug"
] | [] | 2026-04-28T22:17:57Z | 2026-05-03T06:50:42Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | Gabiii2398 | 279,440,179 | U_kgDOEKfrMw | User | false |
openclaw/openclaw | 4,346,601,541 | I_kwDOQb6kR88AAAABAxPgRQ | 73,837 | https://github.com/openclaw/openclaw/issues/73837 | https://api.github.com/repos/openclaw/openclaw/issues/73837 | [Bug]: Node.js auto-installer fails silently with ioctl errors then falsely reports success before crashing | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
When running ./install.sh manually on WSL2 Ubuntu 24.04 (because of this previously stated [Issue#73814](https://github.com/openclaw/openclaw/issues/73814) the Node.js auto-installation step fails silently with ioctl errors, fa... | open | null | false | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-28T22:20:24Z | 2026-05-03T09:49:29Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | ItsMeForLua | 171,505,882 | U_kgDOCjj42g | User | false |
openclaw/openclaw | 4,346,932,223 | I_kwDOQb6kR88AAAABAxjr_w | 73,868 | https://github.com/openclaw/openclaw/issues/73868 | https://api.github.com/repos/openclaw/openclaw/issues/73868 | Server-side timeout for sessions stuck in 'processing' state | Sessions can remain in 'processing' state indefinitely if the agent_end hook fails or a subagent's announce never arrives. The gateway has a [diagnostic] stuck session detector but no auto-failure mechanism. Proposal: configurable timeout (default 2h?) after which a 'processing' session with queueDepth=0 and no recent ... | closed | not_planned | true | 1 | [] | [] | 2026-04-28T23:39:02Z | 2026-05-01T04:41:09Z | 2026-04-28T23:42:57Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dknoodle | 428,382 | MDQ6VXNlcjQyODM4Mg== | User | false |
openclaw/openclaw | 4,346,932,292 | I_kwDOQb6kR88AAAABAxjsRA | 73,869 | https://github.com/openclaw/openclaw/issues/73869 | https://api.github.com/repos/openclaw/openclaw/issues/73869 | agent_end hook completion should reap :cron:run: registry entries | When a cron-driven 'agent:<x>:cron:<id>:run:<run-id>' session completes, the entry persists in agents/<x>/sessions/sessions.json indefinitely. After days of cron firings, registry bloats to thousands of entries that the gateway tries to resume on every restart. Proposal: when agent_end fires for a session whose key mat... | closed | completed | true | 1 | [] | [] | 2026-04-28T23:39:03Z | 2026-05-01T04:41:08Z | 2026-04-28T23:43:06Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dknoodle | 428,382 | MDQ6VXNlcjQyODM4Mg== | User | false |
openclaw/openclaw | 4,346,932,371 | I_kwDOQb6kR88AAAABAxjskw | 73,870 | https://github.com/openclaw/openclaw/issues/73870 | https://api.github.com/repos/openclaw/openclaw/issues/73870 | Per-parent concurrency cap on subagent spawns | Multi-wave fanout patterns (e.g., agent:main spawning 5 subagents in parallel) saturate the gateway's event loop because each subagent's announce callback contends for the same WebSocket. Proposal: configurable per-parent concurrency limit (default 2-3) with subagent spawns queueing past the limit. Observed: 30-agent f... | open | null | false | 1 | [] | [] | 2026-04-28T23:39:05Z | 2026-05-03T06:07:49Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | dknoodle | 428,382 | MDQ6VXNlcjQyODM4Mg== | User | false |
openclaw/openclaw | 4,346,931,382 | I_kwDOQb6kR88AAAABAxjotg | 73,867 | https://github.com/openclaw/openclaw/issues/73867 | https://api.github.com/repos/openclaw/openclaw/issues/73867 | openclaw tasks maintenance --apply should also prune session registry | Currently 'tasks maintenance' only prunes task-flows. With registry cleanup not yet handled by agent_end (see related issue), maintenance should also offer a session registry sweep with conservative criteria (entries with updatedAt >7d AND not currently in jobs-state.json running set). | open | null | false | 3 | [] | [] | 2026-04-28T23:38:48Z | 2026-05-03T07:45:31Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | dknoodle | 428,382 | MDQ6VXNlcjQyODM4Mg== | User | false |
openclaw/openclaw | 4,346,982,729 | I_kwDOQb6kR88AAAABAxmxSQ | 73,874 | https://github.com/openclaw/openclaw/issues/73874 | https://api.github.com/repos/openclaw/openclaw/issues/73874 | Gateway HTTP/WS dispatch deadlock on Windows + Docker Desktop bind-mount setups (regression in 2026.4.24, persists in .25 and .26) | # Gateway HTTP/WS dispatch deadlock on Windows + Docker Desktop bind-mount setups (regression in 2026.4.24, persists in .25 and .26)
## Summary
On Windows host + Docker Desktop + bind-mounted `~/.openclaw/` setups, the gateway in 2026.4.24, 2026.4.25, and 2026.4.26 logs `ready` and binds the listener socket, but the ... | open | null | false | 5 | [] | [] | 2026-04-28T23:52:56Z | 2026-05-03T04:44:04Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | purpleant | 94,114 | MDQ6VXNlcjk0MTE0 | User | false |
openclaw/openclaw | 4,346,646,649 | I_kwDOQb6kR88AAAABAxSQeQ | 73,838 | https://github.com/openclaw/openclaw/issues/73838 | https://api.github.com/repos/openclaw/openclaw/issues/73838 | [Bug]: Control UI intermittently loses webchat state and renders NO_REPLY as NO | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
## Summary
The OpenClaw Control UI/webchat sometimes appears to lose conversation or draft state, making it look like the user’s typed message/conversation disappeared. Separately, assistant responses that are... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-28T22:31:02Z | 2026-05-01T04:41:07Z | 2026-04-28T22:35:10Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | zachisfine | 131,436,334 | U_kgDOB9WPLg | User | false |
openclaw/openclaw | 4,347,037,033 | I_kwDOQb6kR88AAAABAxqFaQ | 73,881 | https://github.com/openclaw/openclaw/issues/73881 | https://api.github.com/repos/openclaw/openclaw/issues/73881 | openclaw infer model run (without --gateway) hangs indefinitely on 4.26; same prompt completes immediately in 4.25 | # Upstream bug report DRAFT — pending Trent blind-pass
**Target:** github.com/openclaw/openclaw/issues
**Status:** DRAFT v0 — not filed.
---
## Title
`openclaw infer model run` hangs on 4.26; child is killed only by caller timeout
## Environment
- OpenClaw `2026.4.26` (be8c246)
- Install: npm-global at `/home/ope... | closed | completed | true | 1 | [] | [] | 2026-04-29T00:08:31Z | 2026-05-01T04:41:02Z | 2026-04-29T00:13:05Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | islandpreneur007 | 258,422,617 | U_kgDOD2c3WQ | User | false |
openclaw/openclaw | 4,347,036,879 | I_kwDOQb6kR88AAAABAxqEzw | 73,879 | https://github.com/openclaw/openclaw/issues/73879 | https://api.github.com/repos/openclaw/openclaw/issues/73879 | Heartbeat / cron / exec wakeups submit empty user prompt to model providers in 4.25, causing 400 INVALID_ARGUMENT on Vertex/Gemini | # Upstream bug report DRAFT — pending Trent blind-pass
**Target:** github.com/openclaw/openclaw/issues
**Status:** DRAFT v0 — not filed. Needs Trent review (RCA primarily lives in his workspace memory).
---
## Title
Heartbeat / cron / exec wakeups submit empty user prompt to model providers in 4.25, causing 400 INV... | closed | completed | true | 1 | [] | [] | 2026-04-29T00:08:27Z | 2026-05-01T04:41:01Z | 2026-04-29T00:13:33Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | islandpreneur007 | 258,422,617 | U_kgDOD2c3WQ | User | false |
openclaw/openclaw | 4,336,943,447 | I_kwDOQb6kR88AAAABAoCBVw | 72,881 | https://github.com/openclaw/openclaw/issues/72881 | https://api.github.com/repos/openclaw/openclaw/issues/72881 | [Bug]: | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Messages sent through Azure Bot Test in Web Chat reach the OpenClaw Microsoft Teams webhook and receive replies, but messages sent through the installed Microsoft Teams app do not produce inbound gateway activi... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-27T15:35:36Z | 2026-05-03T04:36:07Z | 2026-04-30T16:06:25Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | SimSef | 56,121,054 | MDQ6VXNlcjU2MTIxMDU0 | User | false |
openclaw/openclaw | 4,347,156,165 | I_kwDOQb6kR88AAAABAxxWxQ | 73,895 | https://github.com/openclaw/openclaw/issues/73895 | https://api.github.com/repos/openclaw/openclaw/issues/73895 | Gateway v2 metadata: emit promotion-grade routing surface | Gateway v2 metadata: emit promotion-grade routing surface
Problem
- Benchmark-side smoke harness is ready, but OpenClaw gateway still does not emit the full promotion-grade v2 metadata surface.
Needed fields
- requested_config
- resolved_config
- actual_config
- route_class
- fallback_chain
- policy_id
- policy_versi... | open | null | false | 1 | [] | [] | 2026-04-29T00:46:44Z | 2026-05-03T11:50:09Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | faisalcredologi | 262,386,576 | U_kgDOD6OzkA | User | false |
openclaw/openclaw | 4,347,166,529 | I_kwDOQb6kR88AAAABAxx_QQ | 73,900 | https://github.com/openclaw/openclaw/issues/73900 | https://api.github.com/repos/openclaw/openclaw/issues/73900 | [Bug]: main/systemEvent cron heartbeat inherits global heartbeat.to and leaks topic reminder to DM | ## Summary
A `sessionTarget: "main"` + `payload.kind: "systemEvent"` cron job that was bound to a Telegram forum topic session woke the correct topic heartbeat session, but its reminder reply was delivered to the user's Telegram DM instead of the originating topic.
This looks like a core delivery/heartbeat merge bug:... | open | null | false | 1 | [] | [] | 2026-04-29T00:49:21Z | 2026-05-03T10:41:42Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | richardmqq | 14,178,267 | MDQ6VXNlcjE0MTc4MjY3 | User | false |
openclaw/openclaw | 4,347,174,591 | I_kwDOQb6kR88AAAABAxyevw | 73,901 | https://github.com/openclaw/openclaw/issues/73901 | https://api.github.com/repos/openclaw/openclaw/issues/73901 | Feature Request: User approval gate for config.patch beyond allowlist | ## Summary
Currently, `gateway config.patch` enforces a strict allowlist (`ALLOWED_GATEWAY_CONFIG_PATHS`) that only permits changes to low-risk paths (agent model/thinking/prompt, channel requireMention). This was introduced in v2026.4.20 as a fix for GHSA-7jm2-g593-4qrc.
**Problem:** The allowlist is too narrow for ... | open | null | false | 1 | [] | [] | 2026-04-29T00:51:36Z | 2026-05-03T05:00:00Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | RoseKongPS | 270,016,663 | U_kgDOEBgglw | User | false |
openclaw/openclaw | 4,347,196,394 | I_kwDOQb6kR88AAAABAxzz6g | 73,903 | https://github.com/openclaw/openclaw/issues/73903 | https://api.github.com/repos/openclaw/openclaw/issues/73903 | openclaw doctor: Claude CLI section reports irrelevant info for non-Claude-CLI agents | ## Summary
`openclaw doctor` always reports the Claude CLI section against the **default agent**, even when that agent does not use Claude CLI as its backend. The reported `Workspace` and `Claude project dir` paths then look like Claude CLI is bound to that agent, which is misleading.
## Reproduction
Setup:
- Default... | closed | completed | true | 2 | [] | [] | 2026-04-29T00:58:28Z | 2026-05-02T04:29:28Z | 2026-04-29T04:47:53Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | bobfreeman1989 | 49,040,836 | MDQ6VXNlcjQ5MDQwODM2 | User | false |
openclaw/openclaw | 4,347,198,395 | I_kwDOQb6kR88AAAABAxz7uw | 73,905 | https://github.com/openclaw/openclaw/issues/73905 | https://api.github.com/repos/openclaw/openclaw/issues/73905 | Feature: Configurable subagent notify policy for cron jobs | ## Problem
Cron jobs default to `silent` notify policy, but subagents spawned inside cron jobs default to `done_only`. This causes unexpected notifications when cron jobs delegate work to subagents.
**Current behavior:**
| Source | Default notify policy |
|--------|----------------------|
| Cron jobs | `silent` |
| S... | closed | completed | true | 1 | [] | [] | 2026-04-29T00:59:18Z | 2026-05-03T04:34:56Z | 2026-04-30T21:31:07Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | arlen8411 | 28,994,439 | MDQ6VXNlcjI4OTk0NDM5 | User | false |
openclaw/openclaw | 4,347,202,503 | I_kwDOQb6kR88AAAABAx0Lxw | 73,906 | https://github.com/openclaw/openclaw/issues/73906 | https://api.github.com/repos/openclaw/openclaw/issues/73906 | [Bug]: Runtime context message leaked into assistant public reply and persisted to downstream LCM | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Runtime context text appeared in assistant's public reply instead of remaining internal, and downstream LCM persisted it as a normal conversation message.
### Steps to reproduce
NOT_ENOUGH_INFO
### Expected ... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-29T01:00:45Z | 2026-05-01T04:41:00Z | 2026-04-29T01:06:25Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | RayySummers | 106,456,682 | U_kgDOBlhmag | User | false |
openclaw/openclaw | 4,347,079,565 | I_kwDOQb6kR88AAAABAxsrjQ | 73,885 | https://github.com/openclaw/openclaw/issues/73885 | https://api.github.com/repos/openclaw/openclaw/issues/73885 | Feature: Implement Selenium/Puppeteer browser automation | ## Summary\nAdd first-class Selenium/Puppeteer support to OpenClaw for browser automation tasks (web scraping, form filling, automated testing, screenshot capture of complex SPAs).\n\n## Motivation\n- The existing browser tool is lightweight but lacks the control needed for complex web interactions\n- Selenium/Puppetee... | closed | not_planned | true | 1 | [] | [] | 2026-04-29T00:22:45Z | 2026-05-01T04:40:59Z | 2026-04-29T01:07:06Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | roxoctoclaw | 270,063,312 | U_kgDOEBjW0A | User | false |
openclaw/openclaw | 4,347,037,109 | I_kwDOQb6kR88AAAABAxqFtQ | 73,882 | https://github.com/openclaw/openclaw/issues/73882 | https://api.github.com/repos/openclaw/openclaw/issues/73882 | 4.26 doctor surfaces memory-readiness 'not checked' warning on every run; previous behavior (4.25) only warned on actual probe timeout | # Upstream bug report
**Target:** github.com/openclaw/openclaw/issues
---
## Title
4.26 doctor surfaces memory-readiness "not checked" warning on every run; previous behavior (4.25) only warned on actual probe timeout
## Environment
- OpenClaw `2026.4.26` (be8c246)
- Memory provider: ollama (configured)
- Install... | closed | completed | true | 2 | [] | [] | 2026-04-29T00:08:33Z | 2026-05-02T04:26:18Z | 2026-04-30T04:10:52Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | islandpreneur007 | 258,422,617 | U_kgDOD2c3WQ | User | false |
openclaw/openclaw | 4,347,244,059 | I_kwDOQb6kR88AAAABAx2uGw | 73,910 | https://github.com/openclaw/openclaw/issues/73910 | https://api.github.com/repos/openclaw/openclaw/issues/73910 | BUG: OpenClaw-managed Codex ACP uses isolated CODEX_HOME without auth bridge and sends unsupported timeout config | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
OpenClaw-managed Codex ACP sessions fail even though direct ACPX to Codex succeeds. The managed path prepares an isolated Codex home without Codex auth, and after auth is bridged locally it still sends an unsup... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-29T01:15:15Z | 2026-05-03T06:07:51Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,347,268,883 | I_kwDOQb6kR88AAAABAx4PEw | 73,913 | https://github.com/openclaw/openclaw/issues/73913 | https://api.github.com/repos/openclaw/openclaw/issues/73913 | [Bug]: WhatsApp inbound handler dead after 408 reconnect cycle — watchdog detects but reconnect doesn't restore (2026.4.26 regression) | ## Summary
After upgrading to 2026.4.26 (which included WhatsApp stability improvements), the WhatsApp inbound handler enters a silent failure loop within <24 hours. The built-in watchdog correctly detects the problem (`app-silent`) but the 499 reconnect it triggers does not restore inbound message delivery. This pers... | closed | not_planned | true | 1 | [] | [] | 2026-04-29T01:24:38Z | 2026-05-01T04:40:57Z | 2026-04-29T01:24:59Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | primecoder-legonhill | 265,703,021 | U_kgDOD9ZObQ | User | false |
openclaw/openclaw | 4,347,279,143 | I_kwDOQb6kR88AAAABAx43Jw | 73,918 | https://github.com/openclaw/openclaw/issues/73918 | https://api.github.com/repos/openclaw/openclaw/issues/73918 | Config-reload deferral logged but not honored — systemd SIGTERM kills gateway, in-flight user message lost with no retry | # Config-reload deferral honored, but systemd SIGTERM kills gateway 9s later — inbound user message dropped, no retry
## Summary
When a config change triggers a gateway restart, the gateway's "defer until N operations complete" logic runs and logs the deferral, but **systemd issues SIGTERM ~9 seconds later regardless... | closed | completed | true | 1 | [] | [] | 2026-04-29T01:28:40Z | 2026-05-03T04:36:29Z | 2026-04-30T14:59:49Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | shandutta | 8,773,397 | MDQ6VXNlcjg3NzMzOTc= | User | false |
openclaw/openclaw | 4,347,292,646 | I_kwDOQb6kR88AAAABAx5r5g | 73,924 | https://github.com/openclaw/openclaw/issues/73924 | https://api.github.com/repos/openclaw/openclaw/issues/73924 | [Bug]: Gateway hangs from session cleanup 2026.4.x | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
After 2026.4.x the gateway hangs and session cleanup writes are spammed to console.
https://pastebin.com/jCc0pVHJ
https://pastebin.com/xTtva5XD
https://pastebin.com/3h6TbxGE
Disabling Active-Memory and Dreaming made a very sli... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-29T01:34:00Z | 2026-05-01T04:40:55Z | 2026-04-29T01:41:51Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | keough99 | 13,394,623 | MDQ6VXNlcjEzMzk0NjIz | User | false |
openclaw/openclaw | 4,347,291,389 | I_kwDOQb6kR88AAAABAx5m_Q | 73,922 | https://github.com/openclaw/openclaw/issues/73922 | https://api.github.com/repos/openclaw/openclaw/issues/73922 | Gateway crashes ~18s after startup due to bonjour name conflict with other local services | ## Bug Description
Gateway starts successfully but crashes after ~18 seconds with `unhandled_rejection`, entering crashloop. The root cause is a **bonjour/mDNS name conflict** between the OpenClaw gateway and other local services registered as LaunchAgents on the same machine.
## Environment
- **Hardware:** Mac min... | closed | completed | true | 1 | [] | [] | 2026-04-29T01:33:41Z | 2026-05-01T04:40:54Z | 2026-04-29T01:42:14Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | zj8060999-sudo | 267,206,406 | U_kgDOD-0_Bg | User | false |
openclaw/openclaw | 4,347,301,130 | I_kwDOQb6kR88AAAABAx6NCg | 73,926 | https://github.com/openclaw/openclaw/issues/73926 | https://api.github.com/repos/openclaw/openclaw/issues/73926 | Bug: `/new` produces empty prompt → Anthropic API 400 "messages: at least one message is required" | # Bug: `/new` produces empty prompt → Anthropic API 400 "messages: at least one message is required"
## Summary
When `/new` (or `/reset`) is invoked on an agent, there is a race condition between the active-memory subsystem and the bare-reset prompt injection that can cause `runEmbeddedPiAgent` to be called with an e... | closed | completed | true | 1 | [] | [] | 2026-04-29T01:36:54Z | 2026-05-05T04:27:54Z | 2026-05-02T09:34:52Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | giangthb | 45,944,331 | MDQ6VXNlcjQ1OTQ0MzMx | User | false |
openclaw/openclaw | 4,347,287,963 | I_kwDOQb6kR88AAAABAx5Zmw | 73,921 | https://github.com/openclaw/openclaw/issues/73921 | https://api.github.com/repos/openclaw/openclaw/issues/73921 | infer model run silently ignores --model override (always routes to anthropic/claude-sonnet-4-6) | ## Summary
`openclaw infer model run --model <copilot-model-id>` silently ignores the `--model` flag and routes the request to `anthropic/claude-sonnet-4-6` regardless of which model was specified.
## Repro (observed Apr 29 2026, ~01:10 UTC)
```
openclaw infer model run --model github-copilot/gemini-3-flash-preview --... | open | null | false | 2 | [] | [] | 2026-04-29T01:32:20Z | 2026-05-03T06:07:53Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | xonaman | 45,240,351 | MDQ6VXNlcjQ1MjQwMzUx | User | false |
openclaw/openclaw | 4,347,237,976 | I_kwDOQb6kR88AAAABAx2WWA | 73,908 | https://github.com/openclaw/openclaw/issues/73908 | https://api.github.com/repos/openclaw/openclaw/issues/73908 | [Bug]: Gateway unconditionally probes all 115 manifest providers and external CLI credentials at startup — causing 50s+ blocking even with 1 configured provider | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Gateway startup (sidecars.channels / prewarmConfiguredPrimaryModel) blocks ~56s and credential discovery calls security find-generic-password for Claude on macOS, even with only 1 configured provider, 3 plugins... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-29T01:13:13Z | 2026-05-02T04:29:01Z | 2026-04-29T06:52:54Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | Ailuras | 34,333,040 | MDQ6VXNlcjM0MzMzMDQw | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.