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 | 4,076,637,440 | I_kwDOQb6kR87y_I0A | 46,548 | https://github.com/openclaw/openclaw/issues/46548 | https://api.github.com/repos/openclaw/openclaw/issues/46548 | Tool error messages surfaced to chat should include failure reason and retry confirmation | ## Problem
When a tool call (e.g., file edit) fails, OpenClaw surfaces a message to the chat like:
```
⚠️ 📝 Edit: `in ~/.openclaw/workspace/MEMORY.md (77 chars)` failed
```
Two issues with this:
1. **No failure reason is shown.** The user has no idea why it failed — was it a text mismatch? Permission error? File n... | open | null | false | 3 | [] | [] | 2026-03-14T20:14:49Z | 2026-05-04T02:38:20Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | mculp | 129,983 | MDQ6VXNlcjEyOTk4Mw== | User | false |
openclaw/openclaw | 4,076,713,450 | I_kwDOQb6kR87y_bXq | 46,570 | https://github.com/openclaw/openclaw/issues/46570 | https://api.github.com/repos/openclaw/openclaw/issues/46570 | Memory search never returns results from memory files (only sessions) | ## Steps to reproduce
1. Write content to a memory file: `echo '## Test Entry' >> ~/.openclaw/workspace/memory/test.md`
2. Force reindex: `openclaw memory index --force`
3. Verify indexed: `openclaw memory status` → shows memory files indexed with chunks
4. Search for the content: `openclaw memory search 'Test Entry'`
... | open | null | false | 2 | [] | [] | 2026-03-14T20:55:38Z | 2026-05-04T01:20:06Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | matthewcarano | 30,149,603 | MDQ6VXNlcjMwMTQ5NjAz | User | false |
openclaw/openclaw | 4,076,770,632 | I_kwDOQb6kR87y_pVI | 46,585 | https://github.com/openclaw/openclaw/issues/46585 | https://api.github.com/repos/openclaw/openclaw/issues/46585 | iPhone Continuity Camera `--facing front` always uses rear camera | **Description**
When using iPhone Continuity Camera via OpenClaw nodes camera snap, the `--facing front` parameter does not switch to the front camera. Both `--facing front` and `--facing back` capture photos from the rear camera.
**Environment**
- OpenClaw: 2026.3.13 (61d171a)
- Node: MacBook Pro (Mac14,6), macOS S... | open | null | false | 2 | [] | [] | 2026-03-14T21:23:47Z | 2026-05-04T13:20:16Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | roamer-remote | 267,348,107 | U_kgDOD-9oiw | User | false |
openclaw/openclaw | 4,076,790,969 | I_kwDOQb6kR87y_uS5 | 46,590 | https://github.com/openclaw/openclaw/issues/46590 | https://api.github.com/repos/openclaw/openclaw/issues/46590 | Feature Request: Add `cron` field to Agent configuration for Agent-owned scheduled tasks | ## Problem
Currently, all cron jobs are stored centrally in `~/.openclaw/cron/jobs.json` at the Gateway layer, detached from Agent configuration. While jobs can be bound to an agent via `agentId`, the task ownership and configuration are separated, causing:
1. **Configuration fragmentation**: Agent settings (workspac... | open | null | false | 2 | [] | [] | 2026-03-14T21:33:54Z | 2026-05-04T13:20:18Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | roamer-remote | 267,348,107 | U_kgDOD-9oiw | User | false |
openclaw/openclaw | 4,077,007,817 | I_kwDOQb6kR87zAjPJ | 46,637 | https://github.com/openclaw/openclaw/issues/46637 | https://api.github.com/repos/openclaw/openclaw/issues/46637 | [Bug]:reasoning_content in conversation history causes oMLX JSON parse error on subsequent turns | ### Bug type
Regression (worked before, now fails)
### Summary
When Qwen 3.5 returns `reasoning_content` in the openai-completions response, that content appears to be included in the conversation history for the next API request. The serialized JSON body sent to oMLX contains an invalid control character, causing o... | open | null | false | 7 | [
"bug",
"regression"
] | [] | 2026-03-14T23:11:23Z | 2026-05-04T01:20:08Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | zipzagster | 97,322,703 | U_kgDOBc0Gzw | User | false |
openclaw/openclaw | 4,077,075,191 | I_kwDOQb6kR87zAzr3 | 46,656 | https://github.com/openclaw/openclaw/issues/46656 | https://api.github.com/repos/openclaw/openclaw/issues/46656 | Feature: Webchat / Control UI inline button support (buttons parameter, callback RPC, capabilities config) | ## Feature Request
### Summary
The `message` tool supports a `buttons` parameter for sending inline keyboard buttons, but this only works on Telegram today. Webchat (Control UI) clients silently drop the `buttons` payload — no rendering, no callback mechanism, and no capabilities config path exists for webchat.
This... | open | null | false | 3 | [] | [] | 2026-03-14T23:39:58Z | 2026-05-04T13:20:20Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | sys-fairy-eve | 261,672,760 | U_kgDOD5jPOA | User | false |
openclaw/openclaw | 4,077,246,689 | I_kwDOQb6kR87zBdjh | 46,690 | https://github.com/openclaw/openclaw/issues/46690 | https://api.github.com/repos/openclaw/openclaw/issues/46690 | No guard against skill repos created outside canonical paths | ## Bug Summary
When creating or scaffolding skills (via `package_skill.py`, manual creation, or skill-creator workflows), nothing prevents the skill directory from being created outside the canonical `workspace/skills/` path. This leads to orphaned skill directories in locations like `~/trigger-trap-skill/` or `~/my-s... | open | null | false | 2 | [] | [] | 2026-03-15T01:04:53Z | 2026-05-04T13:20:21Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,077,284,043 | I_kwDOQb6kR87zBmrL | 46,701 | https://github.com/openclaw/openclaw/issues/46701 | https://api.github.com/repos/openclaw/openclaw/issues/46701 | Telegram: Auto-reply to unauthorized group senders with admin contact instructions | ## Feature Request
When a user sends a message in a Telegram group but is not in `groupAllowFrom`, the bot currently silently drops the message.
**Requested behavior:** Instead of silently dropping, the bot should reply to the unauthorized sender with a friendly message explaining they need to be authorized, and ins... | open | null | false | 4 | [] | [] | 2026-03-15T01:24:19Z | 2026-05-04T13:20:23Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | natechadwick-intsof | 263,952,448 | U_kgDOD7uYQA | User | false |
openclaw/openclaw | 4,077,350,447 | I_kwDOQb6kR87zB24v | 46,733 | https://github.com/openclaw/openclaw/issues/46733 | https://api.github.com/repos/openclaw/openclaw/issues/46733 | [Bug]: opus 4.6 1M context now allowed with oauth, broken in openclaw | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Since March 13th `opus-4.6` now supports 1M token window on the Claude MAX plan.
Openclaw strips the header and hard-codes the 200k token limit
### Steps to reproduce
Config:
```json
"agents": {
"defaults": {
"models": {
... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T02:01:10Z | 2026-05-04T13:20:25Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Thinkscape | 270,528 | MDQ6VXNlcjI3MDUyOA== | User | false |
openclaw/openclaw | 4,077,438,554 | I_kwDOQb6kR87zCMZa | 46,776 | https://github.com/openclaw/openclaw/issues/46776 | https://api.github.com/repos/openclaw/openclaw/issues/46776 | [Bug]: Feishu group talk with openclaw agents lost messages | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Bug 1: after I talk with a feishu openclaw agent in a session, than I talk to the same feishu openclaw agent in another group session, I found the feishu openclaw agent did not know what I talked to it in before session.
Bug 2: in the feis... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T03:00:23Z | 2026-05-04T01:32:12Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | RichardQidian | 252,679,334 | U_kgDODw-Upg | User | false |
openclaw/openclaw | 4,077,444,549 | I_kwDOQb6kR87zCN3F | 46,778 | https://github.com/openclaw/openclaw/issues/46778 | https://api.github.com/repos/openclaw/openclaw/issues/46778 | 飞书消息重复推送，message_id 去重失效 | **问题描述**
飞书私聊消息经常重复显示两次，每次推送的 message_id 不同，导致现有的去重机制失效。
**复现方式**
1. 在飞书私聊场景中使用 OpenClaw
2. 发送消息给机器人
3. 部分消息会收到两次（同一条消息，不同的 message_id）
**期望行为**
同一条消息只应该被处理一次，不应该重复显示。
**当前问题**
从 `~/.openclaw/feishu/dedup/default.json` 记录来看：
- 去重机制已启用（基于 message_id）
- 但飞书对同一条消息推送了不同的 message_id
- 导致两条相同的消息都被处理
**建议解决方案**
增加基于消... | open | null | false | 2 | [] | [] | 2026-03-15T03:04:48Z | 2026-05-04T13:20:27Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | kingcuty | 61,933,811 | MDQ6VXNlcjYxOTMzODEx | User | false |
openclaw/openclaw | 4,077,448,304 | I_kwDOQb6kR87zCOxw | 46,780 | https://github.com/openclaw/openclaw/issues/46780 | https://api.github.com/repos/openclaw/openclaw/issues/46780 | Feature Request: Skill Lifecycle Hooks (beforeExecute/afterExecute) | ## Description
Currently, OpenClaw supports hooks for commands like `/new`, `/reset`, `/stop`, but there is no way to intercept skill execution.
## Problem
We want to implement a self-improving skills system that can:
1. Log skill execution results automatically
2. Analyze failure patterns
3. Suggest improvements to... | open | null | false | 3 | [] | [] | 2026-03-15T03:07:22Z | 2026-05-04T13:20:28Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | daddyqian-cpu | 264,820,737 | U_kgDOD8jYAQ | User | false |
openclaw/openclaw | 4,077,459,879 | I_kwDOQb6kR87zCRmn | 46,786 | https://github.com/openclaw/openclaw/issues/46786 | https://api.github.com/repos/openclaw/openclaw/issues/46786 | [Bug]: tools.elevated.enabled: true breaks exec routing logic (tools. | ### Bug type
Regression (worked before, now fails)
### Summary
Enabling `tools.elevated.enabled: true` causes ALL `exec` tool calls to route to the gateway host instead of the sandbox, even when the agent does not pass `elevated: true`. The `ask` and `on` levels behave identically — both route everything to the gate... | open | null | false | 5 | [
"bug",
"regression"
] | [] | 2026-03-15T03:15:14Z | 2026-05-04T02:38:25Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | Thinkscape | 270,528 | MDQ6VXNlcjI3MDUyOA== | User | false |
openclaw/openclaw | 4,077,466,681 | I_kwDOQb6kR87zCTQ5 | 46,797 | https://github.com/openclaw/openclaw/issues/46797 | https://api.github.com/repos/openclaw/openclaw/issues/46797 | Config patch restart notice is misleading; should say 'restart required' | When config.patch is used, the gateway posts a system notice like "Gateway restart config-patch ok (config.patch)" followed by the note. This reads as if the restart already happened, but on Windows we explicitly require a manual restart (restart-gateway.ps1) after approval. Please adjust the system notice wording to ... | open | null | false | 2 | [] | [] | 2026-03-15T03:20:07Z | 2026-05-04T13:20:30Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Stache73 | 7,196,384 | MDQ6VXNlcjcxOTYzODQ= | User | false |
openclaw/openclaw | 4,077,474,604 | I_kwDOQb6kR87zCVMs | 46,809 | https://github.com/openclaw/openclaw/issues/46809 | https://api.github.com/repos/openclaw/openclaw/issues/46809 | Control UI: Memory panel for MEMORY.md and USER.md | ## Feature Request
### Summary
Add a dedicated memory panel in the Control UI that surfaces the agent's long-term memory files (MEMORY.md, USER.md) as readable and editable content.
### Motivation
Agent memory is currently a black box to the user unless they open a file editor or ask the agent to read it aloud. This ... | open | null | false | 3 | [] | [] | 2026-03-15T03:25:42Z | 2026-05-04T13:20:31Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | allisonshawai-ctrl | 262,603,479 | U_kgDOD6cC1w | User | false |
openclaw/openclaw | 4,077,529,103 | I_kwDOQb6kR87zCigP | 46,844 | https://github.com/openclaw/openclaw/issues/46844 | https://api.github.com/repos/openclaw/openclaw/issues/46844 | Feature: Talk Mode Idle Timeout / Auto-Deactivation After Voice Wake | ## Summary
After Voice Wake activates Talk Mode, the pipeline stays active indefinitely (up to the 120s hard stop). There is no configurable **idle timeout** that deactivates the pipeline when the user stops interacting. This causes unnecessary token consumption and API calls when the user is done talking but hasn't e... | open | null | false | 3 | [] | [] | 2026-03-15T04:06:07Z | 2026-05-04T13:20:33Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | reidmcgill | 207,939,519 | U_kgDODGTnvw | User | false |
openclaw/openclaw | 4,077,601,765 | I_kwDOQb6kR87zC0Pl | 46,886 | https://github.com/openclaw/openclaw/issues/46886 | https://api.github.com/repos/openclaw/openclaw/issues/46886 | cron wake action does not support agentId — always routes to default agent | ## Summary
The `cron` tool's `wake` action does not accept an `agentId` parameter, causing all wake events to be routed to the default agent (typically `main`) regardless of which agent should be woken.
## Current Behavior
The `wake` function in the cron service only accepts `text` and `mode`:
```js
// gateway-cli ... | open | null | false | 2 | [] | [] | 2026-03-15T04:49:29Z | 2026-05-04T13:20:35Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | aidyfeng | 13,725,371 | MDQ6VXNlcjEzNzI1Mzcx | User | false |
openclaw/openclaw | 4,077,611,969 | I_kwDOQb6kR87zC2vB | 46,894 | https://github.com/openclaw/openclaw/issues/46894 | https://api.github.com/repos/openclaw/openclaw/issues/46894 | [Bug]: Slash command error——WebUI treats `/model status` as a model name instead of a slash-command subcommand | ### Bug type
Regression (worked before, now fails)
### Summary
## Bug Description
In the WebUI, entering `/model status` and `/model list` fails with an error such as:
`Failed to set model: GatewayRequestError: model not allowed: openai-codex/status`
Instead of treating `status` as a request to show current model... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-03-15T04:58:28Z | 2026-05-04T01:32:13Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | xiaoHEI-312 | 44,249,876 | MDQ6VXNlcjQ0MjQ5ODc2 | User | false |
openclaw/openclaw | 4,077,671,100 | I_kwDOQb6kR87zDFK8 | 46,930 | https://github.com/openclaw/openclaw/issues/46930 | https://api.github.com/repos/openclaw/openclaw/issues/46930 | 多 Gateway 检测提示应使用 warn/info 而非 error 级别 | ## 问题描述
当系统检测到多个 gateway 服务运行时，日志输出如下：
```
Other gateway-like services detected (best effort):
- openclaw-family-gateway.service
Recommendation: run a single gateway per machine for most setups.
...
```
这些信息在日志中被标记为 `error` 级别。
## 建议改进
这只是一个**信息提示/建议**，而非真正的错误。多 gateway 配置在某些场景下是合理的（如多 agent 实例隔离）。
建议将日志级别从 `err... | open | null | false | 2 | [] | [] | 2026-03-15T05:43:41Z | 2026-05-04T13:20:36Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | tociyou1995 | 150,312,743 | U_kgDOCPWXJw | User | false |
openclaw/openclaw | 4,077,787,928 | I_kwDOQb6kR87zDhsY | 47,002 | https://github.com/openclaw/openclaw/issues/47002 | https://api.github.com/repos/openclaw/openclaw/issues/47002 | [Bug]: Config validator rejects mediaLocalRoots despite type definitions supporting it (Telegram outbound) | ## Bug type
Config validation / Channels / Telegram / Outbound media
## Summary
`mediaLocalRoots` is defined in the TypeScript types for channel config and agent defaults, and the Telegram extension code fully supports forwarding it — but the Zod config validator rejects it as an "Unrecognized key" at both `channels.t... | open | null | false | 2 | [] | [] | 2026-03-15T07:04:24Z | 2026-05-04T13:20:38Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | dwaynehelena | 18,387,137 | MDQ6VXNlcjE4Mzg3MTM3 | User | false |
openclaw/openclaw | 4,077,885,188 | I_kwDOQb6kR87zD5cE | 47,070 | https://github.com/openclaw/openclaw/issues/47070 | https://api.github.com/repos/openclaw/openclaw/issues/47070 | Attach input provenance to sessions_spawn child tasks | ## Problem
Several inter-session paths already carry provenance, but spawned subagent tasks should also receive explicit lineage on their input. Without that, a spawned child task can look too much like clean first-party input instead of agent-originated inter-session input.
This matters for multi-agent workflows whe... | open | null | false | 2 | [] | [] | 2026-03-15T08:08:21Z | 2026-05-04T13:23:11Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Christoffer91 | 92,569,289 | U_kgDOBYR-yQ | User | false |
openclaw/openclaw | 4,077,930,512 | I_kwDOQb6kR87zEEgQ | 47,095 | https://github.com/openclaw/openclaw/issues/47095 | https://api.github.com/repos/openclaw/openclaw/issues/47095 | Discord voice transcript runs ignore per-channel systemPrompt override | ### Bug
Discord per-channel `systemPrompt` (channel override) is applied for **text** messages but not for **Discord voice (realtime) transcript** turns.
### Environment
- OpenClaw: v2026.3.13
- Channel: Discord voice channel (STT → agent response)
### Repro
1. Configure a per-channel override:
```js
channels: {
d... | closed | completed | true | 4 | [] | [] | 2026-03-15T08:36:19Z | 2026-05-04T04:34:42Z | 2026-05-02T01:49:21Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | qearlyao | 201,136,044 | U_kgDOC_0XrA | User | false |
openclaw/openclaw | 4,077,999,036 | I_kwDOQb6kR87zEVO8 | 47,131 | https://github.com/openclaw/openclaw/issues/47131 | https://api.github.com/repos/openclaw/openclaw/issues/47131 | Feature: inbound 'reply in thread' / 'rit' user directive | ## Summary
Add a built-in inbound message directive that lets users trigger threaded replies by appending phrases like "reply in thread", "rit", "thread", or "in thread" to their message.
## Problem
Currently, threaded replies are controlled by the LLM via `[[reply_to_current]]` tags in its output. This works when c... | open | null | false | 2 | [] | [] | 2026-03-15T09:16:47Z | 2026-05-04T18:21:32Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | andrearzumanyan | 27,296,199 | MDQ6VXNlcjI3Mjk2MTk5 | User | false |
openclaw/openclaw | 4,078,024,351 | I_kwDOQb6kR87zEbaf | 47,143 | https://github.com/openclaw/openclaw/issues/47143 | https://api.github.com/repos/openclaw/openclaw/issues/47143 | memoryFlush never triggers when compactionCount and memoryFlushCompactionCount are both 0 | ## Description
`memoryFlush` never triggers for sessions that have never been compacted, because `hasAlreadyFlushedForCurrentCompaction` incorrectly returns `true` when both counters are 0.
## Root Cause
In `hasAlreadyFlushedForCurrentCompaction`:
```javascript
function hasAlreadyFlushedForCurrentCompaction(entry) ... | open | null | false | 4 | [] | [] | 2026-03-15T09:31:58Z | 2026-05-04T01:32:26Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | JameelHao | 232,899,707 | U_kgDODeHEew | User | false |
openclaw/openclaw | 4,078,128,447 | I_kwDOQb6kR87zE00_ | 47,203 | https://github.com/openclaw/openclaw/issues/47203 | https://api.github.com/repos/openclaw/openclaw/issues/47203 | [Bug]: browser navigate always times out (20s) on Linux VNC — tool timeout races Playwright page.goto load event | ## Summary
`browser navigate` always times out with "Can't reach the OpenClaw browser control service (timed out after 20000ms)" on Linux VPS with VNC display, even for simple sites like baidu.com. All other browser tool actions (`start`, `stop`, `tabs`, `screenshot`, `snapshot`) work perfectly.
**Key finding:** The ... | open | null | false | 3 | [] | [] | 2026-03-15T10:41:28Z | 2026-05-04T20:13:59Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ldfandian | 5,018,331 | MDQ6VXNlcjUwMTgzMzE= | User | false |
openclaw/openclaw | 4,078,181,831 | I_kwDOQb6kR87zFB3H | 47,229 | https://github.com/openclaw/openclaw/issues/47229 | https://api.github.com/repos/openclaw/openclaw/issues/47229 | [Bug]: Anthropic SDK path missing large-integer precision guard (tool_use input) | ## Summary
The `quoteUnsafeIntegerLiterals` fix from #23170 only covers Ollama/llama.cpp JSON parsing paths. The Anthropic Messages API streaming path has the same precision loss bug for tool_use `input` arguments containing large integers (e.g., Discord snowflake IDs).
## Root Cause
When streaming Anthropic respons... | open | null | false | 2 | [] | [] | 2026-03-15T11:18:15Z | 2026-05-04T18:21:34Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | abczsl520 | 65,114,094 | MDQ6VXNlcjY1MTE0MDk0 | User | false |
openclaw/openclaw | 4,078,190,506 | I_kwDOQb6kR87zFD-q | 47,235 | https://github.com/openclaw/openclaw/issues/47235 | https://api.github.com/repos/openclaw/openclaw/issues/47235 | bug: `memory status --deep` embedding readiness probe can flap and disagree with actual index state | ## Summary
After fixing the separate proxy-routing issue for remote embeddings, I still see a different problem in `openclaw memory status --deep`:
- repeated runs for the same agent can alternate between:
- `Embeddings: ready`
- `Embeddings: unavailable`
- `Embeddings error: fetch failed`
- meanwhile, the on-d... | open | null | false | 3 | [] | [] | 2026-03-15T11:24:14Z | 2026-05-04T20:14:01Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ningfeemic-dev | 239,259,264 | U_kgDODkLOgA | User | false |
openclaw/openclaw | 4,078,243,980 | I_kwDOQb6kR87zFRCM | 47,273 | https://github.com/openclaw/openclaw/issues/47273 | https://api.github.com/repos/openclaw/openclaw/issues/47273 | [Bug]: Memory detection skips macOS (darwin) — os.freemem() / os.totalmem() never reported to agent on non-Linux platforms | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
On macOS (darwin), memory detection is gated behind a platform === "linux" check and silently skips, so the agent never receives accurate system memory data.
### Steps to reproduce
1. Run OpenClaw gateway on macOS or a Lume/macOS VM (plat... | open | null | false | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T12:01:42Z | 2026-05-02T18:30:58Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | g466562644-sys | 266,909,281 | U_kgDOD-i2YQ | User | false |
openclaw/openclaw | 4,078,276,398 | I_kwDOQb6kR87zFY8u | 47,294 | https://github.com/openclaw/openclaw/issues/47294 | https://api.github.com/repos/openclaw/openclaw/issues/47294 | feishu_create_doc: LaTeX backslashes eaten in complex block formulas ($$ blocks) | ## Bug Description
When using `feishu_create_doc` (or `feishu_update_doc`) to create Feishu documents containing LaTeX block formulas (`$$...$$`), **backslashes in complex formulas are silently stripped**, causing rendering failures.
## Reproduction
Simple formulas work fine:
```markdown
$$
\frac{a}{b} = c
$$
```
✅ ... | open | null | false | 2 | [] | [] | 2026-03-15T12:23:24Z | 2026-05-02T18:30:59Z | null | NONE | null | 20260502T210030Z | 2026-05-02T21:00:30Z | gaohongkui | 38,203,359 | MDQ6VXNlcjM4MjAzMzU5 | User | false |
openclaw/openclaw | 4,078,306,403 | I_kwDOQb6kR87zFgRj | 47,311 | https://github.com/openclaw/openclaw/issues/47311 | https://api.github.com/repos/openclaw/openclaw/issues/47311 | [Feature Request] Add server-side STT option for webchat voice input | ## Summary
Add a server-side speech-to-text option for webchat, allowing voice input to be processed by local Whisper (or other configured STT backends) instead of relying on browser's Web Speech API.
## Problem
The current webchat voice input uses the browser's native `SpeechRecognition` API (Web Speech API), which... | closed | completed | true | 3 | [] | [] | 2026-03-15T12:43:17Z | 2026-05-05T04:26:06Z | 2026-05-02T23:09:25Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | jmomford | 17,583,828 | MDQ6VXNlcjE3NTgzODI4 | User | false |
openclaw/openclaw | 4,078,319,845 | I_kwDOQb6kR87zFjjl | 47,320 | https://github.com/openclaw/openclaw/issues/47320 | https://api.github.com/repos/openclaw/openclaw/issues/47320 | Feature Request: recursive subagent listing (--depth / --recursive) | ## Problem
With `maxSpawnDepth: 2`, we use the orchestrator pattern: main agent -> project manager (depth-1) -> worker experts (depth-2). The main agent needs visibility into depth-2 workers to monitor project progress, but `subagents list` only shows direct children.
## Current behavior
- `subagents list` returns d... | open | null | false | 3 | [] | [] | 2026-03-15T12:52:29Z | 2026-05-02T15:22:27Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | shibiao1998-dot | 266,466,674 | U_kgDOD-H1cg | User | false |
openclaw/openclaw | 4,078,360,424 | I_kwDOQb6kR87zFtdo | 47,335 | https://github.com/openclaw/openclaw/issues/47335 | https://api.github.com/repos/openclaw/openclaw/issues/47335 | Reply generated but not delivered when compaction triggers session rollover | ## Bug Description
When a session approaches context limits (~85% usage) and auto-compaction triggers with `willRetry=true` (overflow recovery), the LLM generates a reply that is visible in the session transcript (`stopReason: "stop"`), but the reply is never delivered to the messaging channel (Discord, Telegram, etc.... | open | null | false | 4 | [] | [] | 2026-03-15T13:18:31Z | 2026-05-02T14:21:22Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | frankhli843 | 51,806,469 | MDQ6VXNlcjUxODA2NDY5 | User | false |
openclaw/openclaw | 4,078,369,368 | I_kwDOQb6kR87zFvpY | 47,342 | https://github.com/openclaw/openclaw/issues/47342 | https://api.github.com/repos/openclaw/openclaw/issues/47342 | Node host: no SSH transport support + challenge auth fails behind reverse proxy | ## Summary
`openclaw node run` cannot connect to a remote gateway behind a TLS reverse proxy (nginx). The gateway's built-in SSH transport (`gateway.remote.transport: "ssh"`) works for the gateway client but is **not available for the node host**, forcing users to manage SSH tunnels externally.
Additionally, when con... | open | null | false | 2 | [] | [] | 2026-03-15T13:24:11Z | 2026-05-02T16:55:00Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | crhan | 532,040 | MDQ6VXNlcjUzMjA0MA== | User | false |
openclaw/openclaw | 4,078,463,906 | I_kwDOQb6kR87zGGui | 47,382 | https://github.com/openclaw/openclaw/issues/47382 | https://api.github.com/repos/openclaw/openclaw/issues/47382 | Feature: Channel Silence Watchdog — runtime beacon during long model requests | ## Summary
Add a configurable silence watchdog to the gateway channel delivery layer. When no operator-visible output has been delivered for a configurable duration during active work, the gateway automatically sends a brief status message to the channel.
## Problem
During long reasoning blocks (no tool calls, no st... | closed | not_planned | false | 2 | [] | [] | 2026-03-15T14:14:59Z | 2026-05-04T16:02:34Z | 2026-05-04T16:02:34Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | CaiusPrime | 262,411,762 | U_kgDOD6QV8g | User | false |
openclaw/openclaw | 4,078,624,941 | I_kwDOQb6kR87zGuCt | 47,441 | https://github.com/openclaw/openclaw/issues/47441 | https://api.github.com/repos/openclaw/openclaw/issues/47441 | sharp image module fails on Android/Termux after every update — no android-arm64 prebuilt, NDK cross-compiler assumed | ## Environment
- Device: Pixel 8 Pro, Android 14
- Runtime: Termux + proot-distro (glibc), Node v25.3.0
- Architecture: aarch64 (android-arm64)
- OpenClaw running as headless gateway with Telegram provider
## Problem
After every `openclaw update`, the `image` tool fails with:
```
Could not load the "sharp" module usin... | open | null | false | 3 | [] | [] | 2026-03-15T15:46:03Z | 2026-05-02T16:40:21Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | rocketpoweryul | 5,898,307 | MDQ6VXNlcjU4OTgzMDc= | User | false |
openclaw/openclaw | 4,078,646,974 | I_kwDOQb6kR87zGza- | 47,452 | https://github.com/openclaw/openclaw/issues/47452 | https://api.github.com/repos/openclaw/openclaw/issues/47452 | [Bug]: HTTP 400 error when using moonshotai/kimi-k2.5: Invalid value 'thinking' not supported by Kimi API | ### Bug type
Regression (worked before, now fails)
### Summary
HTTP 400 error when using moonshotai/kimi-k2.5: Invalid value 'thinking' not supported by Kimi API
### Steps to reproduce
1. Configure OpenClaw to use moonshotai/kimi-k2.5 as the primary model
2. Set API type to anthropic-messages
3. Start any con... | open | null | false | 4 | [
"bug",
"regression"
] | [] | 2026-03-15T15:59:27Z | 2026-05-02T17:30:24Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | gitking | 10,758,854 | MDQ6VXNlcjEwNzU4ODU0 | User | false |
openclaw/openclaw | 4,078,726,004 | I_kwDOQb6kR87zHGt0 | 47,487 | https://github.com/openclaw/openclaw/issues/47487 | https://api.github.com/repos/openclaw/openclaw/issues/47487 | [Bug]: Agent tool profile restrictions not enforced - exec available despite not in alsoAllow list | ### Bug type
Regression (worked before, now fails)
### Summary
Bug: When an agent has tools.profile: "messaging" with a specific alsoAllow list that does NOT include exec, the exec tool is still available.
### Steps to reproduce
Steps to reproduce:
Set global config tools.exec: { security: "full", ask: "off" ... | closed | completed | true | 6 | [
"bug",
"regression"
] | [] | 2026-03-15T16:42:18Z | 2026-05-03T04:37:37Z | 2026-04-30T12:19:28Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | bennetkoch | 263,215,705 | U_kgDOD7BaWQ | User | false |
openclaw/openclaw | 4,078,736,476 | I_kwDOQb6kR87zHJRc | 47,492 | https://github.com/openclaw/openclaw/issues/47492 | https://api.github.com/repos/openclaw/openclaw/issues/47492 | feat(exec): verified-human instruction provenance for sensitive targets | ## Summary
Add a separate, gateway-enforced policy layer for **instruction provenance** on sensitive side-effect actions.
CaMeL / taint provenance addresses the question:
- **Did untrusted content influence this action?**
This feature request addresses a different question:
- **Did a verified human actually reques... | open | null | false | 2 | [] | [] | 2026-03-15T16:48:11Z | 2026-05-02T17:30:26Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | BearHuddleston | 196,404,565 | U_kgDOC7TlVQ | User | false |
openclaw/openclaw | 4,078,796,783 | I_kwDOQb6kR87zHX_v | 47,516 | https://github.com/openclaw/openclaw/issues/47516 | https://api.github.com/repos/openclaw/openclaw/issues/47516 | Discord: Reaction listeners registered unconditionally even when all guilds have reactionNotifications=off | ## Summary
Reaction event listeners (DiscordReactionListener, DiscordReactionRemoveListener) are always registered regardless of guild configuration. When all guilds have `reactionNotifications: "off"`, reaction events still hit the event queue, go through guild config resolution and authorization checks, then return ... | closed | completed | true | 3 | [] | [] | 2026-03-15T17:20:36Z | 2026-05-04T04:34:05Z | 2026-05-02T03:56:28Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | x4v13r1120 | 54,415,425 | MDQ6VXNlcjU0NDE1NDI1 | User | false |
openclaw/openclaw | 4,078,840,915 | I_kwDOQb6kR87zHixT | 47,534 | https://github.com/openclaw/openclaw/issues/47534 | https://api.github.com/repos/openclaw/openclaw/issues/47534 | [Feature] doctor detects stale session snapshots after runtime-path changes | ## Summary
`openclaw doctor` should detect **stale session snapshots / cached metadata** that still reference old runtime roots after OpenClaw has been moved, rewired, upgraded, or temporarily run from the wrong path.
This bit me through bundled skill paths, but the problem is broader than one skill.
## Concrete fai... | open | null | false | 5 | [] | [] | 2026-03-15T17:44:29Z | 2026-05-04T01:32:58Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | sene1337 | 261,265,124 | U_kgDOD5KW5A | User | false |
openclaw/openclaw | 4,078,894,604 | I_kwDOQb6kR87zHv4M | 47,555 | https://github.com/openclaw/openclaw/issues/47555 | https://api.github.com/repos/openclaw/openclaw/issues/47555 | [Bug]: --profile flag ignored when executing gateway restart from agent session | ## Description
When executing `openclaw --profile <profile-name> gateway restart` from within an agent session (e.g., `agent:main`), the `--profile` flag is silently ignored and the command operates on the current session's gateway instead of the specified profile.
## Reproduction Steps
1. Start a main gateway sessi... | open | null | false | 4 | [] | [] | 2026-03-15T18:11:43Z | 2026-05-04T01:33:02Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | caelvorin | 246,398,649 | U_kgDODq--uQ | User | false |
openclaw/openclaw | 4,078,905,821 | I_kwDOQb6kR87zHynd | 47,558 | https://github.com/openclaw/openclaw/issues/47558 | https://api.github.com/repos/openclaw/openclaw/issues/47558 | docs(subagents): 6 gaps/inaccuracies in sub-agents documentation | Hey — went through `docs/tools/subagents.md` carefully against the actual source and found 6 issues. Some are just missing info, a couple are genuinely misleading.
### 1. Missing `announceTimeoutMs` in config reference
**Severity:** Medium
`agents.defaults.subagents.announceTimeoutMs` (default: 90000ms) exists in the ... | closed | completed | true | 4 | [] | [] | 2026-03-15T18:16:59Z | 2026-05-04T04:34:58Z | 2026-05-01T23:20:12Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | akrimm702 | 144,416,483 | U_kgDOCJue4w | User | false |
openclaw/openclaw | 4,078,920,228 | I_kwDOQb6kR87zH2Ik | 47,565 | https://github.com/openclaw/openclaw/issues/47565 | https://api.github.com/repos/openclaw/openclaw/issues/47565 | Feature: Configurable Permissions-Policy and CSP headers for Control UI | ## Problem
The gateway hardcodes security headers for the Control UI static file serving:
```
Permissions-Policy: camera=(), microphone=(), geolocation=()
Content-Security-Policy: ... connect-src 'self' ws: wss: ...
```
These are set in `gateway-cli-*.js` and cannot be configured via `openclaw.json`.
This blocks le... | open | null | false | 3 | [] | [] | 2026-03-15T18:24:01Z | 2026-05-04T01:33:05Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | spacexengineer | 6,034,810 | MDQ6VXNlcjYwMzQ4MTA= | User | false |
openclaw/openclaw | 4,078,986,462 | I_kwDOQb6kR87zIGTe | 47,584 | https://github.com/openclaw/openclaw/issues/47584 | https://api.github.com/repos/openclaw/openclaw/issues/47584 | feat(ios): Siri App Intent integration — native "Hey Siri, ask OpenClaw..." support | ## Summary
The iOS app currently lacks native Siri App Intent support. Users can work around this with Shortcuts + Webhook, but there's no deep integration where Siri can handle parameters, stream replies, or initiate a conversation natively.
This issue proposes adding `AppIntent` conformance to the iOS app so users ... | open | null | false | 2 | [] | [] | 2026-03-15T19:01:26Z | 2026-05-04T01:22:00Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | peteradams2026 | 259,958,501 | U_kgDOD36m5Q | User | false |
openclaw/openclaw | 4,079,008,184 | I_kwDOQb6kR87zILm4 | 47,586 | https://github.com/openclaw/openclaw/issues/47586 | https://api.github.com/repos/openclaw/openclaw/issues/47586 | Reminder guard fires false positives on memory-related "I'll remember" statements | ## Description
The `agent-runner-reminder-guard` appends a visible disclaimer to agent responses when the model says things like "I'll remember" without scheduling a cron job. However, the regex pattern for "remember" is too broad — it matches memory/knowledge retention statements, not just temporal reminder commitmen... | open | null | false | 3 | [] | [] | 2026-03-15T19:13:57Z | 2026-05-04T01:33:06Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | moltpill | 261,749,793 | U_kgDOD5n8IQ | User | false |
openclaw/openclaw | 4,079,009,241 | I_kwDOQb6kR87zIL3Z | 47,587 | https://github.com/openclaw/openclaw/issues/47587 | https://api.github.com/repos/openclaw/openclaw/issues/47587 | Bug: Telegram multi-image messages - first image transcribed but excluded from media header, others in media header but not transcribed | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When sending multiple images in a single Telegram message, OpenClaw only transcribes the first image and excludes its path from the media attached header. Other images are included in the header but not transcribed.
### Steps to reproduce
... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-15T19:14:33Z | 2026-05-04T01:33:07Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | yzjJosh | 14,024,301 | MDQ6VXNlcjE0MDI0MzAx | User | false |
openclaw/openclaw | 4,079,040,713 | I_kwDOQb6kR87zITjJ | 47,596 | https://github.com/openclaw/openclaw/issues/47596 | https://api.github.com/repos/openclaw/openclaw/issues/47596 | [Feature Request] Add tool call visibility to streamTo="parent" for ACP runtime | # [Feature Request] Add tool call visibility to streamTo="parent" for ACP runtime
## Current Behavior
When using `runtime="acp"` with `streamTo="parent"`, only the following events are streamed back to the parent session:
1. **Assistant text output** - Text deltas from the child agent
2. **Lifecycle events** - Start... | open | null | false | 2 | [] | [] | 2026-03-15T19:33:28Z | 2026-05-04T01:33:11Z | null | CONTRIBUTOR | null | 20260504T030031Z | 2026-05-04T03:00:31Z | uf-hy | 41,638,541 | MDQ6VXNlcjQxNjM4NTQx | User | false |
openclaw/openclaw | 4,079,040,889 | I_kwDOQb6kR87zITl5 | 47,597 | https://github.com/openclaw/openclaw/issues/47597 | https://api.github.com/repos/openclaw/openclaw/issues/47597 | [Feature Request] Add streamTo="parent" support for runtime="subagent" | # [Feature Request] Add streamTo="parent" support for runtime="subagent"
## Current Behavior
The `streamTo="parent"` parameter is currently **only supported for `runtime="acp"`**:
```javascript
sessions_spawn({
runtime: "subagent",
streamTo: "parent" // ❌ Error: streamTo is only supported for runtime=acp
})
```... | open | null | false | 5 | [] | [] | 2026-03-15T19:33:34Z | 2026-05-04T01:33:12Z | null | CONTRIBUTOR | null | 20260504T030031Z | 2026-05-04T03:00:31Z | uf-hy | 41,638,541 | MDQ6VXNlcjQxNjM4NTQx | User | false |
openclaw/openclaw | 4,079,079,040 | I_kwDOQb6kR87zIc6A | 47,606 | https://github.com/openclaw/openclaw/issues/47606 | https://api.github.com/repos/openclaw/openclaw/issues/47606 | Execution anti-drift guards: artifact-gated status, escalation timers, and no-idle-after-green | ## Problem
OpenClaw agent sessions can enter a status-loop/escalation-drift pattern during multi-batch execution:
- repeated status updates without execution artifacts,
- delayed blocker escalation,
- idle gaps after CI green before next-step execution.
This wastes operator time and reduces reliability.
## Proposal (... | open | null | false | 2 | [] | [] | 2026-03-15T19:56:57Z | 2026-05-04T01:33:15Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | Claudebert | 263,088,221 | U_kgDOD65oXQ | User | false |
openclaw/openclaw | 4,079,283,886 | I_kwDOQb6kR87zJO6u | 47,651 | https://github.com/openclaw/openclaw/issues/47651 | https://api.github.com/repos/openclaw/openclaw/issues/47651 | Feature request: Multimodal memory indexing via Gemini Embedding 2 | ## Feature Request
### What
Support indexing of non-Markdown media files (images, video, audio, PDFs) in the memory search pipeline using Gemini Embedding 2's native multimodal capabilities.
### Why
Gemini Embedding 2 (`gemini-embedding-2-preview`) maps text, images, video, audio, and PDFs into a single unified vecto... | open | null | false | 3 | [] | [] | 2026-03-15T21:59:19Z | 2026-05-04T01:33:21Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | Zak-neurobit | 211,141,578 | U_kgDODJXDyg | User | false |
openclaw/openclaw | 4,079,459,600 | I_kwDOQb6kR87zJ50Q | 47,690 | https://github.com/openclaw/openclaw/issues/47690 | https://api.github.com/repos/openclaw/openclaw/issues/47690 | Feature: multi-account routing for bot-to-bot communication in Telegram groups | ---
date: 2026-03-16
agent: Anubis
topic: telegram, multi-agent, limitation
title: Limitación: bots no se ven entre sí en grupos de Telegram
---
## Problema
En un grupo de Telegram con múltiples bots (Vision, Skynet, Anubis), cada bot solo ve sus **propios mensajes**. No ven los mensajes enviados por los otros bots.
... | open | null | false | 3 | [] | [] | 2026-03-15T23:35:12Z | 2026-05-04T01:33:24Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | lunaviva211-sketch | 262,210,777 | U_kgDOD6EE2Q | User | false |
openclaw/openclaw | 4,079,602,870 | I_kwDOQb6kR87zKcy2 | 47,712 | https://github.com/openclaw/openclaw/issues/47712 | https://api.github.com/repos/openclaw/openclaw/issues/47712 | TTS filterEmoji 配置建议 | ## 建议
在 TTS 工具中添加 filterEmoji 选项，默认过滤 emoji 后再转语音。
## 背景
用户希望 TTS 语音生成时跳过 emoji，避免 emoji 被朗读出来。
## 现状
已在配置文件中添加选项：
```json
"messages": { "tts": { "filterEmoji": true } }
```
但 TTS 工具尚未支持读取此配置。
## 期望
希望 TTS 工具能原生支持自动过滤 emoji，提升语音质量。 | open | null | false | 3 | [] | [] | 2026-03-16T00:40:00Z | 2026-05-04T01:33:28Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | Robinyst | 35,587,042 | MDQ6VXNlcjM1NTg3MDQy | User | false |
openclaw/openclaw | 4,079,856,697 | I_kwDOQb6kR87zLaw5 | 47,812 | https://github.com/openclaw/openclaw/issues/47812 | https://api.github.com/repos/openclaw/openclaw/issues/47812 | Feature: Smart session switch reminder based on context usage and topic change | ## Summary
Add intelligent session switch reminders when context is running low and the user switches topics.
## Proposed Behavior
| Condition | Action |
|-----------|--------|
| Context ≥80% + topic switch detected | Prompt: "Want to start a new session?" |
| Context ≥95% | Auto-switch to new session |
## Topic Sw... | open | null | false | 2 | [] | [] | 2026-03-16T02:26:08Z | 2026-05-01T23:59:26Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | xuhe83-cyber | 247,283,756 | U_kgDODr1ALA | User | false |
openclaw/openclaw | 4,079,902,071 | I_kwDOQb6kR87zLl13 | 47,826 | https://github.com/openclaw/openclaw/issues/47826 | https://api.github.com/repos/openclaw/openclaw/issues/47826 | Gateway should expose HTTP REST endpoints for common queries | ## Problem
`openclaw status --json` takes **30-40 seconds** on modest hardware (Intel i7-9750H, Node v22) because the CLI does a full cold-start: transpiles TypeScript, loads 30+ provider plugins, connects via WebSocket, queries data. The actual gateway response is <50ms — the other 99% is CLI bootstrap overhead.
Thi... | open | null | false | 2 | [] | [] | 2026-03-16T02:45:26Z | 2026-05-02T01:29:08Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | hurttlocker | 195,353,125 | U_kgDOC6TaJQ | User | false |
openclaw/openclaw | 4,079,939,910 | I_kwDOQb6kR87zLvFG | 47,834 | https://github.com/openclaw/openclaw/issues/47834 | https://api.github.com/repos/openclaw/openclaw/issues/47834 | browser tool: include current page URL in before_tool_call params for act/type actions | ## Problem
When a plugin hook runs on `before_tool_call` for the `browser` tool with `act`/`type` actions, the params don't include the browser's current page URL. Only `ref`, `text`, `kind`, etc. are present.
This is needed for domain-pinning security checks — the credential vault plugin needs to verify the current ... | open | null | false | 4 | [] | [] | 2026-03-16T03:01:03Z | 2026-05-03T04:43:12Z | null | CONTRIBUTOR | null | 20260503T060030Z | 2026-05-03T06:00:30Z | karanuppal | 1,457,892 | MDQ6VXNlcjE0NTc4OTI= | User | false |
openclaw/openclaw | 4,079,961,377 | I_kwDOQb6kR87zL0Uh | 47,840 | https://github.com/openclaw/openclaw/issues/47840 | https://api.github.com/repos/openclaw/openclaw/issues/47840 | Control UI: Model provider display shows original config value instead of normalized value (e.g., 'bailian' vs 'zai') | ### Describe the bug
The Control UI displays the **original configuration value** for model providers instead of the **normalized value** used internally, causing confusion when switching between providers.
### To Reproduce
1. Configure a model in openclaw.json with provider 'bailian' (which is an alias for 'zai')
2... | open | null | false | 4 | [] | [] | 2026-03-16T03:09:35Z | 2026-05-01T23:59:27Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | def177 | 139,129,111 | U_kgDOCErxFw | User | false |
openclaw/openclaw | 4,080,049,656 | I_kwDOQb6kR87zMJ34 | 47,856 | https://github.com/openclaw/openclaw/issues/47856 | https://api.github.com/repos/openclaw/openclaw/issues/47856 | Feature: configurable mediaLocalRoots for image tool | ## Problem
The `image` tool only allows reads from a fixed set of directories (`~/.openclaw/workspace/`, sandbox paths). There is no way to configure additional allowed roots.
This is a problem for iMessage users with `includeAttachments: true`. Attachment paths arrive as something like:
```
/Users/<user>/Library/Me... | open | null | false | 5 | [] | [] | 2026-03-16T03:47:23Z | 2026-05-01T23:59:29Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | chad42777 | 266,836,478 | U_kgDOD-eZ_g | User | false |
openclaw/openclaw | 4,080,242,611 | I_kwDOQb6kR87zM4-z | 47,910 | https://github.com/openclaw/openclaw/issues/47910 | https://api.github.com/repos/openclaw/openclaw/issues/47910 | feat: provider fallback by failure class — quarantine auth-broken providers | ## Problem
Current model failover treats all failures equally — auth failure, rate limit, network timeout, and invalid request all trigger the same fallback chain. This wastes latency retrying providers with known-bad auth states.
## Real-world impact
From logs:
```
candidate_failed ... reason=auth next=openai-code... | open | null | false | 5 | [] | [] | 2026-03-16T05:02:57Z | 2026-05-04T20:14:04Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | issaba1 | 9,679,850 | MDQ6VXNlcjk2Nzk4NTA= | User | false |
openclaw/openclaw | 4,080,243,575 | I_kwDOQb6kR87zM5N3 | 47,911 | https://github.com/openclaw/openclaw/issues/47911 | https://api.github.com/repos/openclaw/openclaw/issues/47911 | feat: openclaw fleet status — unified multi-agent observability command | ## Problem
Debugging a multi-agent deployment requires grepping `journalctl` across multiple services, manually checking auth profiles, inspecting session states, and cross-referencing channel configs. There's no single command to answer "what's the state of my fleet?"
## Real-world impact
Running 8 agents across se... | open | null | false | 2 | [] | [] | 2026-03-16T05:03:15Z | 2026-05-01T23:59:31Z | null | CONTRIBUTOR | null | 20260502T000037Z | 2026-05-02T00:00:37Z | issaba1 | 9,679,850 | MDQ6VXNlcjk2Nzk4NTA= | User | false |
openclaw/openclaw | 4,080,430,849 | I_kwDOQb6kR87zNm8B | 47,940 | https://github.com/openclaw/openclaw/issues/47940 | https://api.github.com/repos/openclaw/openclaw/issues/47940 | Heartbeat alternates between sent and ok-token every cycle — effective interval is 2x configured | ## Environment
- OpenClaw 2026.3.13, macOS, GPT-5.4
- `heartbeat.every`: 30m
- `heartbeat.target`: whatsapp
- `lightContext`: false
## Observed behavior
Heartbeat fires every 30 minutes as configured, but alternates between successful exploration and silent ok-token:
```
10:21 ✅ sent (diary written, WhatsApp se... | open | null | false | 13 | [] | [] | 2026-03-16T06:05:53Z | 2026-05-03T03:22:06Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | Leolai6-7 | 112,496,602 | U_kgDOBrSP2g | User | false |
openclaw/openclaw | 4,080,552,863 | I_kwDOQb6kR87zOEuf | 47,964 | https://github.com/openclaw/openclaw/issues/47964 | https://api.github.com/repos/openclaw/openclaw/issues/47964 | Possible auth profile compatibility issue for openai-codex OAuth credentials | ## Observed behavior
In one local deployment, `openai-codex` OAuth login appeared to complete successfully and credentials were written to `auth-profiles.json`, but runtime still behaved as if no usable credential was available.
## Symptom
The runtime kept falling back as though Codex auth was unavailable until the ... | open | null | false | 5 | [] | [] | 2026-03-16T06:39:10Z | 2026-05-02T17:30:28Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | firstdraft-work | 265,695,765 | U_kgDOD9YyFQ | User | false |
openclaw/openclaw | 4,080,664,579 | I_kwDOQb6kR87zOgAD | 47,983 | https://github.com/openclaw/openclaw/issues/47983 | https://api.github.com/repos/openclaw/openclaw/issues/47983 | [Bug] WebChat 上传按钮点击无反应 | ## 问题描述
在 OpenClaw 3.13 版本的 WebChat 控制面板中，对话框有上传按钮，但点击上传按钮后没有任何反应。
## 环境信息
- OpenClaw 版本：2026.3.13
- 浏览器：Chrome
- 操作系统：Windows
- Channel：WebChat
## 预期行为
点击上传按钮应该：
1. 打开文件选择对话框
2. 选择文件后可以将文件作为附件发送
## 实际行为
点击上传按钮后没有任何反应，文件选择对话框没有打开。
## 代码分析
根据源代码分析，上传按钮的 HTML 和点击事件都存在：
- class="agent-chat__file-input"
- @click 事件绑定... | open | null | false | 3 | [] | [] | 2026-03-16T07:09:51Z | 2026-05-01T23:59:32Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | Fission21 | 80,576,421 | MDQ6VXNlcjgwNTc2NDIx | User | false |
openclaw/openclaw | 4,081,022,336 | I_kwDOQb6kR87zP3WA | 48,045 | https://github.com/openclaw/openclaw/issues/48045 | https://api.github.com/repos/openclaw/openclaw/issues/48045 | [Bug]: Browser tool silently discards downloads and throws "Download is starting" error when using CDP connection | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When Playwright connects via CDP (connectOverCDP), file downloads are silently discarded, and navigating to URLs with Content-Disposition: attachment throws an unhandled "Download is starting" error instead of returning a success response.
... | open | null | false | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-16T08:34:48Z | 2026-05-02T03:02:29Z | null | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | roinou532 | 207,696,752 | U_kgDODGEzcA | User | false |
openclaw/openclaw | 4,081,406,259 | I_kwDOQb6kR87zRVEz | 48,108 | https://github.com/openclaw/openclaw/issues/48108 | https://api.github.com/repos/openclaw/openclaw/issues/48108 | Control UI sessions dropdown shows deleted sessions from cache | ## Bug Description
**Steps to reproduce:**
1. Open Control UI (http://127.0.0.1:18789/)
2. Select a Discord session from the sessions dropdown
3. Delete the session using `/new` or `sessions.patch`
4. The session is deleted from backend (`sessions.json`)
5. But the session still appears in the dropdown menu on the UI
... | open | null | false | 3 | [] | [] | 2026-03-16T09:51:25Z | 2026-05-01T23:59:34Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | caochen223 | 128,443,466 | U_kgDOB6fkSg | User | false |
openclaw/openclaw | 4,081,460,490 | I_kwDOQb6kR87zRiUK | 48,117 | https://github.com/openclaw/openclaw/issues/48117 | https://api.github.com/repos/openclaw/openclaw/issues/48117 | [Bug]: npm error code UNABLE_TO_VERIFT_LEAF_SIGNATURE | ### Bug type
Regression (worked before, now fails)
### Summary
Fail to install openclaw in Powershell with `iwr -useb https://openclaw.ai/install.ps1 | iex`
### Steps to reproduce
just excute `iwr -useb https://openclaw.ai/install.ps1 | iex`
### Expected behavior
install successfully
### Actual behavior
Failed... | open | null | false | 3 | [
"bug",
"regression"
] | [] | 2026-03-16T10:00:53Z | 2026-05-01T22:57:48Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | Scientist-Ma | 42,742,667 | MDQ6VXNlcjQyNzQyNjY3 | User | false |
openclaw/openclaw | 4,082,074,614 | I_kwDOQb6kR87zT4P2 | 48,183 | https://github.com/openclaw/openclaw/issues/48183 | https://api.github.com/repos/openclaw/openclaw/issues/48183 | [Bug]: Feishu monitor state cleanup incomplete - potential memory leak in httpServers Map | ## Summary
The Feishu plugin's monitor state cleanup in `monitor.state.ts` has a potential memory leak. When stopping the monitor, the httpServers are closed but the Map entries are deleted immediately without waiting for the server to fully close.
## Steps to Reproduce
1. Configure OpenClaw with multiple Feishu acc... | open | null | false | 16 | [] | [] | 2026-03-16T12:05:45Z | 2026-05-03T02:14:31Z | null | NONE | null | 20260503T030030Z | 2026-05-03T03:00:30Z | ai-nurmamat | 130,913,840 | U_kgDOB82WMA | User | false |
openclaw/openclaw | 4,082,552,709 | I_kwDOQb6kR87zVs-F | 48,238 | https://github.com/openclaw/openclaw/issues/48238 | https://api.github.com/repos/openclaw/openclaw/issues/48238 | Add loop-aware compaction guard for saturated sessions | # GitHub Issue Draft — Loop-Aware Compaction Guard
## Title
Add loop-aware compaction guard for saturated sessions
## Summary
Long-lived chat sessions can degrade before or around compaction when the transcript accumulates repeated tool-failure loops, stale reminder/system drift, and near-duplicate assistant turns. W... | open | null | false | 3 | [] | [] | 2026-03-16T13:38:53Z | 2026-05-03T03:36:47Z | null | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | xz0831 | 214,744,049 | U_kgDODMy78Q | User | false |
openclaw/openclaw | 4,084,135,042 | I_kwDOQb6kR87zbvSC | 48,432 | https://github.com/openclaw/openclaw/issues/48432 | https://api.github.com/repos/openclaw/openclaw/issues/48432 | [Bug]: TUI: Code blocks render spurious spaces at line-wrap positions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
In the TUI (openclaw tui), code blocks containing long lines have spurious spaces inserted mid-word at line-wrap positions.
### Steps to reproduce
Display a code block with a line longer than 32 chars.
### Expected behavior
Long lines i... | closed | completed | false | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-16T18:37:04Z | 2026-05-04T13:50:43Z | 2026-05-04T13:50:43Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | DocOellerson | 77,577,906 | MDQ6VXNlcjc3NTc3OTA2 | User | false |
openclaw/openclaw | 4,084,762,209 | I_kwDOQb6kR87zeIZh | 48,486 | https://github.com/openclaw/openclaw/issues/48486 | https://api.github.com/repos/openclaw/openclaw/issues/48486 | [Security] web-fetch-visibility: near-invisible CSS patterns not covered by sanitizer | # [Security] web-fetch-visibility: near-invisible CSS patterns not covered by sanitizer
## Description
The existing `sanitizeHtml()` in `web-fetch-visibility.ts` does an excellent job covering most invisible content injection vectors (HTML comments, `display:none`, `visibility:hidden`, `opacity:0`, `font-size:0`, `co... | open | null | false | 4 | [] | [] | 2026-03-16T20:58:52Z | 2026-05-01T22:57:50Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | lolisaigao1234 | 92,764,750 | U_kgDOBYd6Tg | User | false |
openclaw/openclaw | 4,084,978,791 | I_kwDOQb6kR87ze9Rn | 48,506 | https://github.com/openclaw/openclaw/issues/48506 | https://api.github.com/repos/openclaw/openclaw/issues/48506 | [Feature]: Enrich message_sending hook with resolved target context |
This is the second of three related feature requests for enriching plugin hooks with security-relevant context. The motivation came from implementing CaMeL style prompt injection defense, but each proposal has standalone value. See #48503 for the full background.
### Problem
The message_sending hook lets plugins o... | open | null | false | 2 | [] | [] | 2026-03-16T21:54:48Z | 2026-05-02T01:29:11Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | nashborges | 9,395,917 | MDQ6VXNlcjkzOTU5MTc= | User | false |
openclaw/openclaw | 4,084,995,716 | I_kwDOQb6kR87zfBaE | 48,510 | https://github.com/openclaw/openclaw/issues/48510 | https://api.github.com/repos/openclaw/openclaw/issues/48510 | [Feature]: Add optional capability descriptors to tool definitions |
### Problem
Plugins that need to make policy decisions about tool calls currently rely on tool name matching. This is brittle: tool names change, custom tools have unpredictable names, and the same policy logic ("block tools that send data externally") has to maintain a hardcoded list of tool names.
### Proposal
... | open | null | false | 3 | [] | [] | 2026-03-16T21:58:57Z | 2026-05-02T01:29:12Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | nashborges | 9,395,917 | MDQ6VXNlcjkzOTU5MTc= | User | false |
openclaw/openclaw | 4,085,926,173 | I_kwDOQb6kR87zikkd | 48,684 | https://github.com/openclaw/openclaw/issues/48684 | https://api.github.com/repos/openclaw/openclaw/issues/48684 | Bare abort triggers (stop, abort, wait) delayed/swallowed by inbound debounce | ## Bug
Bare abort phrases like `stop`, `abort`, `wait` are delayed (and can be swallowed entirely) by the inbound message debounce layer on Discord, Slack, Telegram, and other channels with debouncing enabled.
## Root Cause
`shouldDebounceTextInbound()` in `src/channels/inbound-debounce-policy.ts` calls `hasControlC... | open | null | false | 4 | [] | [] | 2026-03-17T03:03:37Z | 2026-05-01T23:59:38Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | sherly-techtree | 258,044,850 | U_kgDOD2Fzsg | User | false |
openclaw/openclaw | 4,086,042,094 | I_kwDOQb6kR87zjA3u | 48,709 | https://github.com/openclaw/openclaw/issues/48709 | https://api.github.com/repos/openclaw/openclaw/issues/48709 | Gemini 2.5 Pro: textSignature bloat + think tags + mixed text/tool causes session failures | ## Summary
When using `google/gemini-2.5-pro` via the `google-generative-ai` API, three issues combine to cause rapid session context bloat, aborted runs, and silent Telegram delivery failures.
## Issue 1: textSignature Session Bloat (High Severity)
Gemini assistant responses include a `textSignature` field containi... | open | null | false | 4 | [] | [] | 2026-03-17T03:51:07Z | 2026-05-01T22:57:51Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | jarvis-playrockets | 261,354,837 | U_kgDOD5P1VQ | User | false |
openclaw/openclaw | 4,086,486,481 | I_kwDOQb6kR87zktXR | 48,780 | https://github.com/openclaw/openclaw/issues/48780 | https://api.github.com/repos/openclaw/openclaw/issues/48780 | [Bug]: [Windows] exec() and read() commands corrupted with </arg_value>> suffix | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
## Description
All `exec()` and `read()` tool calls are being corrupted with `</arg_value>>` suffix on Windows.
### Environment
- **OS**: Windows 10 (NT 10.0.19041 x64)
- **Shell**: PowerShell
- **Node.js**: v25.2.1
- **Platform**: Windo... | open | null | false | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-17T06:23:05Z | 2026-05-02T01:14:31Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | koden588-blip | 265,863,668 | U_kgDOD9jB9A | User | false |
openclaw/openclaw | 4,086,505,920 | I_kwDOQb6kR87zkyHA | 48,785 | https://github.com/openclaw/openclaw/issues/48785 | https://api.github.com/repos/openclaw/openclaw/issues/48785 | [Feature]: Granular session visibility — split tools.sessions.visibility into per-capability knobs |
---
## Summary
Replace the single `tools.sessions.visibility` enum with a structured object of per-capability flags, so deployments can allow cross-agent message delivery without simultaneously granting transcript read access.
---
## Problem to solve
`tools.sessions.visibility` is a single string enum (`"self" | ... | open | null | false | 3 | [] | [] | 2026-03-17T06:28:33Z | 2026-05-01T22:57:52Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | mantisai-bot | 259,690,819 | U_kgDOD3qRQw | User | false |
openclaw/openclaw | 4,086,512,269 | I_kwDOQb6kR87zkzqN | 48,786 | https://github.com/openclaw/openclaw/issues/48786 | https://api.github.com/repos/openclaw/openclaw/issues/48786 | Feishu: replied/quoted message mentions show as raw @_user_N placeholders | ## Bug: Feishu replied/quoted message mentions show as raw `@_user_N` placeholders
### Description
When a Feishu message is a reply to another message that contains `@` mentions, the quoted message body displays raw Feishu mention placeholders (e.g. `@_user_1`) instead of resolving them to `<at user_id="...">name</at... | open | null | false | 5 | [] | [] | 2026-03-17T06:30:25Z | 2026-05-01T22:57:54Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | Raven-wy | 267,524,157 | U_kgDOD_IYPQ | User | false |
openclaw/openclaw | 4,086,564,354 | I_kwDOQb6kR87zlAYC | 48,793 | https://github.com/openclaw/openclaw/issues/48793 | https://api.github.com/repos/openclaw/openclaw/issues/48793 | feat: centralized PluginResourceManager interface for consistent async cleanup across all channel plugins | ## Context
PR #48588 fixes async server cleanup for Feishu, but the same pattern (sync server.close() without awaiting) likely exists in other channel plugins (Telegram, Slack, Discord, etc.). Each plugin implements cleanup differently, leading to recurring port leak and memory leak bugs.
## Proposal
Introduce a Plu... | open | null | false | 3 | [] | [] | 2026-03-17T06:45:04Z | 2026-05-01T23:59:40Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | alex-xuweilong | 267,811,734 | U_kgDOD_Z7lg | User | false |
openclaw/openclaw | 4,086,698,813 | I_kwDOQb6kR87zlhM9 | 48,814 | https://github.com/openclaw/openclaw/issues/48814 | https://api.github.com/repos/openclaw/openclaw/issues/48814 | Feature Request: Pre-send queue check — suppress stale replies when newer messages are pending | ## Problem
When a user sends a message and the agent is generating a response, the user may send a follow-up message that corrects, cancels, or supersedes the original request. Currently, the agent completes its response to the **first** message and sends it, then processes the second message in a new turn.
**Example... | open | null | false | 6 | [] | [] | 2026-03-17T07:21:02Z | 2026-05-01T22:57:55Z | null | NONE | null | 20260502T000037Z | 2026-05-02T00:00:37Z | reconlabs-kyle | 78,457,126 | MDQ6VXNlcjc4NDU3MTI2 | User | false |
openclaw/openclaw | 4,087,533,178 | I_kwDOQb6kR87zos56 | 48,920 | https://github.com/openclaw/openclaw/issues/48920 | https://api.github.com/repos/openclaw/openclaw/issues/48920 | [Bug]: Live Docs are ahead of release | ### Bug type
Regression (worked before, now fails)
### Summary
Heartbeat IsolatedSessions is in the live docs but not in the latest version 2026.3.13
### Steps to reproduce
Add `IsolatedSessions` to your heartbeat config as per the docs:
https://docs.openclaw.ai/gateway/heartbeat#quick-start-beginner
You will ge... | open | null | false | 6 | [
"bug",
"regression"
] | [] | 2026-03-17T10:14:14Z | 2026-05-02T02:11:40Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | Stoff81 | 1,501,310 | MDQ6VXNlcjE1MDEzMTA= | User | false |
openclaw/openclaw | 4,088,936,505 | I_kwDOQb6kR87zuDg5 | 49,062 | https://github.com/openclaw/openclaw/issues/49062 | https://api.github.com/repos/openclaw/openclaw/issues/49062 | [Bug]: Clarify Telegram channel_post auth when groups[chatId].allowFrom is configured | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Telegram native commands on `channel_post` now work for explicitly allowed chats, but sender-scoped per-chat overrides such as
`groups[chatId].allowFrom` remain ambiguous on this surface.
Telegram does not expose the originating human send... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-17T14:27:50Z | 2026-05-02T01:14:33Z | null | NONE | null | 20260502T030029Z | 2026-05-02T03:00:29Z | sonim1 | 2,497,849 | MDQ6VXNlcjI0OTc4NDk= | User | false |
openclaw/openclaw | 4,089,318,284 | I_kwDOQb6kR87zvguM | 49,099 | https://github.com/openclaw/openclaw/issues/49099 | https://api.github.com/repos/openclaw/openclaw/issues/49099 | Extract GIF metadata (source URL, title) from inbound WhatsApp media | ## Problem
When a user sends a GIF from WhatsApp's built-in GIF picker (Giphy/Tenor), the bot only sees `<media:video>` as the message body. The bot has no way to know what the GIF is about without video processing (ffmpeg), which isn't available in the sandbox.
However, WhatsApp GIFs from the picker come with rich m... | open | null | false | 3 | [] | [] | 2026-03-17T15:28:42Z | 2026-05-01T22:57:57Z | null | CONTRIBUTOR | null | 20260502T000037Z | 2026-05-02T00:00:37Z | bugkill3r | 2,924,124 | MDQ6VXNlcjI5MjQxMjQ= | User | false |
openclaw/openclaw | 4,096,512,625 | I_kwDOQb6kR870K9Jx | 49,931 | https://github.com/openclaw/openclaw/issues/49931 | https://api.github.com/repos/openclaw/openclaw/issues/49931 | feat(exec): configurable shell override (tools.exec.shell) | ## Problem
On Windows, the exec tool hardcodes PowerShell as the shell. There's no config option to override this. For agents that need to run complex shell commands with pipes, `jq`, nested quoting, regex, etc., PowerShell's quoting rules make this extremely painful - especially when the command involves JSON process... | open | null | false | 4 | [] | [] | 2026-03-18T17:02:42Z | 2026-05-04T18:21:48Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | jakepresent | 31,285,242 | MDQ6VXNlcjMxMjg1MjQy | User | false |
openclaw/openclaw | 4,096,750,746 | I_kwDOQb6kR870L3Sa | 49,944 | https://github.com/openclaw/openclaw/issues/49944 | https://api.github.com/repos/openclaw/openclaw/issues/49944 | Control UI: tool cards remain collapsed even with verboseDefault=full | ## Summary
Tool cards in Control UI remain collapsed by default even when `agents.defaults.verboseDefault` is set to `full`.
## Environment
- OpenClaw version: `2026.3.13`
- Runtime verbose mode: `full` (confirmed via `session_status`)
- Surface: `webchat` / Control UI
- Browser hard refresh performed
## Reproduction... | open | null | false | 6 | [] | [] | 2026-03-18T17:43:01Z | 2026-05-04T20:14:23Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | RacecarGuy | 13,582,683 | MDQ6VXNlcjEzNTgyNjgz | User | false |
openclaw/openclaw | 4,096,803,543 | I_kwDOQb6kR870MELX | 49,949 | https://github.com/openclaw/openclaw/issues/49949 | https://api.github.com/repos/openclaw/openclaw/issues/49949 | Feature Request: Add Anthropic native web search (web_search_20260209) as a search provider | ## Feature Request
### Summary
Add support for Anthropic's native web search tool (`web_search_20260209` / `web_search_20250305`) as a search provider option alongside the existing Brave, Gemini, Grok, Kimi, and Perplexity providers.
### Background
Anthropic now offers a built-in web search tool that runs **server-... | open | null | false | 2 | [] | [] | 2026-03-18T17:51:58Z | 2026-05-04T18:21:52Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | metasal1 | 54,984,459 | MDQ6VXNlcjU0OTg0NDU5 | User | false |
openclaw/openclaw | 4,097,572,968 | I_kwDOQb6kR870PABo | 50,011 | https://github.com/openclaw/openclaw/issues/50011 | https://api.github.com/repos/openclaw/openclaw/issues/50011 | [Bug]: gateway.nodes.denyCommands silently accepts invalid command names without warning | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
`gateway.nodes.denyCommands` accepts any string without validating against the actual command registry. Invalid command names (typos, wrong format) are silently ignored, giving users a false sense of security.
## To reproduce
1. Set `gatewa... | open | null | false | 3 | [] | [] | 2026-03-18T20:35:27Z | 2026-05-04T18:21:54Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | leonardofhy | 55,347,088 | MDQ6VXNlcjU1MzQ3MDg4 | User | false |
openclaw/openclaw | 4,097,970,725 | I_kwDOQb6kR870QhIl | 50,040 | https://github.com/openclaw/openclaw/issues/50040 | https://api.github.com/repos/openclaw/openclaw/issues/50040 | Telegram delivery reliability: polling stalls can lead to silent outbound message loss | # Telegram delivery reliability: polling stalls can lead to silent outbound message loss
## Summary
On OpenClaw 2026.3.12, Telegram Bot API connectivity may remain generally healthy while the gateway's Telegram polling loop intermittently stalls/restarts. During those recovery windows, outbound `sendMessage` delivery ... | open | null | false | 7 | [] | [] | 2026-03-18T22:08:32Z | 2026-05-04T20:22:36Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | dmitriiforpost-commits | 262,919,577 | U_kgDOD6vVmQ | User | false |
openclaw/openclaw | 4,098,377,268 | I_kwDOQb6kR870SEY0 | 50,081 | https://github.com/openclaw/openclaw/issues/50081 | https://api.github.com/repos/openclaw/openclaw/issues/50081 | Bug: current-run HEIC prompt images reach OpenAI as image/heic instead of being normalized before model send | ## Summary
HEIC/HEIF images uploaded in the OpenClaw Control UI can fail on the **current run** with OpenAI vision models even though the persisted message is later normalized to JPEG successfully.
The user-visible error is:
> The image data you provided does not represent a valid image. Please check your input and ... | open | null | false | 4 | [] | [] | 2026-03-18T23:59:41Z | 2026-05-04T20:22:41Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | kevinheneveld | 1,192,102 | MDQ6VXNlcjExOTIxMDI= | User | false |
openclaw/openclaw | 4,098,431,429 | I_kwDOQb6kR870SRnF | 50,090 | https://github.com/openclaw/openclaw/issues/50090 | https://api.github.com/repos/openclaw/openclaw/issues/50090 | Community Skill Development & ClawHub | Skills are how OpenClaw grows beyond what ships in the box. The promise: write a `SKILL.md`, publish to ClawHub, and any operator can install capabilities you built. A living ecosystem of agent primitives, community-maintained.
The gap between promise and practice is wide right now.
Driftnet has been tracking the ski... | open | null | false | 14 | [] | [] | 2026-03-19T00:16:05Z | 2026-05-04T20:22:42Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | ocdlmv1 | 266,897,137 | U_kgDOD-iG8Q | User | false |
openclaw/openclaw | 4,098,448,434 | I_kwDOQb6kR870SVwy | 50,093 | https://github.com/openclaw/openclaw/issues/50093 | https://api.github.com/repos/openclaw/openclaw/issues/50093 | WhatsApp: Backfill missed messages after reconnection | ## Feature Request
### Problem
When the WhatsApp Web connection drops (e.g. 503 errors) and reconnects, messages sent to monitored groups during the outage window are never delivered or logged. The gateway reconnects successfully but the missed messages are silently lost.
This is particularly impactful for group cha... | open | null | false | 7 | [] | [] | 2026-03-19T00:21:35Z | 2026-05-04T20:22:44Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | maestrokling | 265,137,369 | U_kgDOD82s2Q | User | false |
openclaw/openclaw | 4,098,685,318 | I_kwDOQb6kR870TPmG | 50,126 | https://github.com/openclaw/openclaw/issues/50126 | https://api.github.com/repos/openclaw/openclaw/issues/50126 | Inconsistent message:sent / message_sent hook coverage across multiple outbound dispatch paths | ## Problem Description
OpenClaw currently has **multiple independent outbound message dispatch paths**, but hook triggering coverage is inconsistent. Some outbound messages never trigger `message:sent` (internal) or `message_sent` (plugin) hooks, causing downstream hooks (e.g. conversation-stream) to miss sent messages... | open | null | false | 5 | [] | [] | 2026-03-19T01:39:53Z | 2026-05-04T20:14:38Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Deniable9570 | 266,249,687 | U_kgDOD96l1w | User | false |
openclaw/openclaw | 4,098,861,530 | I_kwDOQb6kR870T6na | 50,145 | https://github.com/openclaw/openclaw/issues/50145 | https://api.github.com/repos/openclaw/openclaw/issues/50145 | interrupt queue mode replays previous assistant reply after abort | ## Bug Description
When `messages.queue.mode` is set to `"interrupt"`, incoming messages that abort an active run cause the **previous turn's assistant reply** to be re-sent to the user, before the new (correct) reply is delivered.
## Steps to Reproduce
1. Set `messages.queue.mode: "interrupt"` in config
2. Send mes... | open | null | false | 5 | [] | [] | 2026-03-19T02:33:43Z | 2026-05-04T20:22:48Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | nano-jason | 246,129,955 | U_kgDODqulIw | User | false |
openclaw/openclaw | 4,098,952,855 | I_kwDOQb6kR870UQ6X | 50,165 | https://github.com/openclaw/openclaw/issues/50165 | https://api.github.com/repos/openclaw/openclaw/issues/50165 | [Bug]: Subagents can appear completed before the underlying delegated work is actually finished | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
OpenClaw subagents can appear finished in the UI/session flow before the underlying delegated work is actually complete, which makes task state unreliable for long-running or multi-step jobs.
## Steps to reproduce
1. Start a long-running or m... | open | null | false | 7 | [] | [] | 2026-03-19T03:00:55Z | 2026-05-04T18:21:59Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | pidgeon777 | 13,739,892 | MDQ6VXNlcjEzNzM5ODky | User | false |
openclaw/openclaw | 4,099,135,680 | I_kwDOQb6kR870U9jA | 50,184 | https://github.com/openclaw/openclaw/issues/50184 | https://api.github.com/repos/openclaw/openclaw/issues/50184 | Telegram DM reply preview is forced to message transport instead of draft | ### Bug type
Regression (worked before, now fails)
### Summary
Telegram direct-message reply streaming regresses after upgrades because the upper reply layer forces DM answer previews to use message transport instead of letting the Telegram draft stream keep its DM-prefers-draft behavior.
### Steps to reproduce
1.... | closed | not_planned | true | 5 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-19T03:56:26Z | 2026-05-02T04:26:40Z | 2026-04-30T02:01:15Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | MuziIsabel | 28,729,467 | MDQ6VXNlcjI4NzI5NDY3 | User | false |
openclaw/openclaw | 4,099,141,904 | I_kwDOQb6kR870U_EQ | 50,186 | https://github.com/openclaw/openclaw/issues/50186 | https://api.github.com/repos/openclaw/openclaw/issues/50186 | [Feature]: Surface systemd gateway hygiene issues in doctor/status and consider safer KillMode defaults | Summary
Improve `openclaw doctor` / `openclaw status --deep` to surface systemd gateway hygiene problems on Linux, and consider a safer default `KillMode` for newly generated gateway units.
Problem to solve
When the Linux user-systemd gateway unit uses `KillMode=process`, restarting the gateway can leave long-lived ... | open | null | false | 4 | [] | [] | 2026-03-19T03:58:14Z | 2026-05-04T20:14:49Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | openclawq | 258,709,591 | U_kgDOD2uYVw | User | false |
openclaw/openclaw | 4,099,179,924 | I_kwDOQb6kR870VIWU | 50,195 | https://github.com/openclaw/openclaw/issues/50195 | https://api.github.com/repos/openclaw/openclaw/issues/50195 | WhatsApp: support pre-built link previews in sendMessage + batch send RPC method | ## Feature Request
### Problem
When sending multiple WhatsApp messages programmatically (e.g., from cron jobs or digest workflows), there are two pain points:
1. **Link previews require per-message HTTP fetches** — Baileys' `generateHighQualityLinkPreview` fetches URL metadata synchronously during each `sendMessage`... | open | null | false | 5 | [] | [] | 2026-03-19T04:12:58Z | 2026-05-04T20:22:56Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Harnoor6693 | 32,019,475 | MDQ6VXNlcjMyMDE5NDc1 | User | false |
openclaw/openclaw | 4,099,221,865 | I_kwDOQb6kR870VSlp | 50,199 | https://github.com/openclaw/openclaw/issues/50199 | https://api.github.com/repos/openclaw/openclaw/issues/50199 | Feature Request: Add Skill Priority Configuration | # Feature Request: Add Skill Priority Configuration
## 📋 Summary
Add support for skill priority configuration to enable intelligent skill selection when multiple skills can perform the same task.
## 🐛 Problem Statement
OpenClaw currently has overlapping skills without clear selection rules, leading to inconsisten... | open | null | false | 7 | [] | [] | 2026-03-19T04:28:10Z | 2026-05-04T20:14:53Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | jimmylzt188 | 69,511,217 | MDQ6VXNlcjY5NTExMjE3 | User | false |
openclaw/openclaw | 4,099,436,304 | I_kwDOQb6kR870WG8Q | 50,245 | https://github.com/openclaw/openclaw/issues/50245 | https://api.github.com/repos/openclaw/openclaw/issues/50245 | health: show gateway probe duration in text output | ## Summary
`openclaw health --json` includes `durationMs`, but the text output from `openclaw health` does not surface the gateway probe duration.
This makes the JSON mode more informative than the default human-facing mode, even though probe timing is useful when diagnosing slow gateway / channel health responses.
... | open | null | false | 3 | [] | [] | 2026-03-19T05:37:45Z | 2026-05-04T18:22:14Z | null | CONTRIBUTOR | null | 20260504T210031Z | 2026-05-04T21:00:31Z | JonathanJing | 17,068,507 | MDQ6VXNlcjE3MDY4NTA3 | User | false |
openclaw/openclaw | 4,099,443,237 | I_kwDOQb6kR870WIol | 50,248 | https://github.com/openclaw/openclaw/issues/50248 | https://api.github.com/repos/openclaw/openclaw/issues/50248 | Bug: sessions cleanup --fix-missing can prune fresh cron sessions with valid transcripts | # OpenClaw bug report — session cleanup falsely prunes fresh cron sessions
Date: 2026-03-18 22:20 America/Phoenix
Reporter: OpenClaw autonomous ops cycle
Project: Mission Control / session-store safety
## Summary
`openclaw sessions cleanup --dry-run --fix-missing` is falsely classifying fresh cron sessions as missing... | open | null | false | 8 | [] | [] | 2026-03-19T05:40:26Z | 2026-05-04T18:22:16Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | lifeofcammy | 214,564,202 | U_kgDODMn9ag | User | false |
openclaw/openclaw | 4,099,627,324 | I_kwDOQb6kR870W1k8 | 50,268 | https://github.com/openclaw/openclaw/issues/50268 | https://api.github.com/repos/openclaw/openclaw/issues/50268 | Bug: Adding a second provider via onboard/configure unconditionally overwrites agents.defaults.model.primary, breaking existing provider (e.g. Bedrock → xAI) | ## Summary
When a user has configured Bedrock as their primary model provider (using `auth: "aws-sdk"` via instance role / env vars) and then adds a second provider via `openclaw onboard` or `openclaw configure` (e.g. xAI, OpenAI), the onboard flow calls `applyAgentDefaultModelPrimary` unconditionally — overwriting `a... | closed | completed | true | 5 | [] | [] | 2026-03-19T06:26:56Z | 2026-05-04T04:34:33Z | 2026-05-02T02:16:57Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | rixcorp-oc | 265,470,397 | U_kgDOD9LBvQ | User | false |
openclaw/openclaw | 4,099,665,276 | I_kwDOQb6kR870W-18 | 50,274 | https://github.com/openclaw/openclaw/issues/50274 | https://api.github.com/repos/openclaw/openclaw/issues/50274 | Follow-up: optimize fallback order + add model-input config guardrail | ## Background
Production fallback expansion รอบล่าสุด deploy สำเร็จแล้ว แต่มีจุดที่ควรทำต่อใน sprint ถัดไปเพื่อให้เสถียรขึ้นและ optimize ต้นทุน/คุณภาพ
## Scope (Next Round)
- [ ] Benchmark + rank fallback models ที่เปิดใช้งานจริงตอนนี้
- openrouter/qwen/qwen3.5-flash-02-23
- openrouter/minimax/minimax-m2.7
- op... | open | null | false | 3 | [] | [] | 2026-03-19T06:37:06Z | 2026-05-04T20:15:00Z | null | NONE | null | 20260504T210031Z | 2026-05-04T21:00:31Z | Piboonsak | 17,107,792 | MDQ6VXNlcjE3MTA3Nzky | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.