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 | 4,114,818,427 | I_kwDOQb6kR871QyV7 | 52,089 | https://github.com/openclaw/openclaw/issues/52089 | https://api.github.com/repos/openclaw/openclaw/issues/52089 | [Bug]: OpenClaw Control UI: Slash Command Menu Position Bug | ### Bug type
Regression (worked before, now fails)
### Summary
## Summary
The slash command menu (`/`) is rendered off-screen (negative `top` position), making it invisible to users.
## Environment
- **OpenClaw Version**: 2026.3.13
- **Browser**: Chrome (latest)
- **OS**: macOS 26.3.1 (arm64)
- **Gateway Mode**: Lo... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-22T05:02:05Z | 2026-04-30T04:49:43Z | 2026-04-25T05:48:59Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Bossiniliu | 142,236,343 | U_kgDOCHpatw | User | false |
openclaw/openclaw | 4,114,832,224 | I_kwDOQb6kR871Q1tg | 52,093 | https://github.com/openclaw/openclaw/issues/52093 | https://api.github.com/repos/openclaw/openclaw/issues/52093 | Windows: EPERM error on atomic write to devices/pending.json (file lock race condition) | ## Bug Description
On Windows, OpenClaw occasionally fails to write `devices/pending.json` with an EPERM error during atomic file rename operation.
## Error Details
**Error Message:**
```
Error: EPERM: operation not permitted, rename 'C:\Users\Administrator\.openclaw\devices\pending.json.{uuid}.tmp' -> 'C:\Users\Adm... | closed | completed | true | 3 | [] | [] | 2026-03-22T05:05:59Z | 2026-04-30T04:49:42Z | 2026-04-25T05:49:06Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cl858085789 | 33,660,289 | MDQ6VXNlcjMzNjYwMjg5 | User | false |
openclaw/openclaw | 4,041,344,778 | I_kwDOQb6kR87w4gcK | 39,901 | https://github.com/openclaw/openclaw/issues/39901 | https://api.github.com/repos/openclaw/openclaw/issues/39901 | [Feature Request] Add option to hide heartbeat poll messages from chat history | ## Description
Currently, heartbeat poll messages are injected into the chat history as system messages. When the agent returns HEARTBEAT_OK, the poll message still appears in the conversation.
This poll message is visible in the chat history and can be distracting.
## Proposed Solution
Add a configuration option t... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-08T13:52:58Z | 2026-04-30T04:49:40Z | 2026-04-25T05:49:38Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lemoncat7 | 59,995,349 | MDQ6VXNlcjU5OTk1MzQ5 | User | false |
openclaw/openclaw | 4,041,613,571 | I_kwDOQb6kR87w5iED | 40,049 | https://github.com/openclaw/openclaw/issues/40049 | https://api.github.com/repos/openclaw/openclaw/issues/40049 | [Backend Feature]: 实现 Append 工具解决 Write 工具数据丢失问题 | ## 背景
接续 #40001 - Write tool lacks append mode
当前问题：isolated cron sessions 使用 write 工具会覆盖共享的 workspace 文件（如 memory/YYYY-MM-DD.md），导致数据丢失。
## 需要实现的功能
### 方案：添加 append 工具
创建一个新的 "append" 工具，只向文件末尾追加内容，不截断文件。
### 工具设计
### 实现位置
1. **创建 append 工具定义**: 在 pi-coding-agent 中或 OpenClaw 包装层
2. **实现文件追加操作**: 使用 fs.append... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-08T16:16:17Z | 2026-04-30T04:49:39Z | 2026-04-25T05:49:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | liangruochong44-ui | 229,361,665 | U_kgDODavIAQ | User | false |
openclaw/openclaw | 4,042,022,483 | I_kwDOQb6kR87w7F5T | 40,206 | https://github.com/openclaw/openclaw/issues/40206 | https://api.github.com/repos/openclaw/openclaw/issues/40206 | Bug: Cron-isolated sessions generate 400+ char tool call IDs, exceeding provider limits | **Bug: Tool call IDs from cron-isolated sessions exceed 64-char limit causing intermittent failures**
### Summary
When cron jobs spawn agents with `sessionTarget: "isolated"`, the generated session keys produce tool call IDs that exceed provider limits (64 chars for Moonshot, 40 for OpenAI). This causes non-determinis... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-08T19:40:27Z | 2026-04-30T04:49:38Z | 2026-04-25T05:49:51Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | seb989241-bot | 258,438,042 | U_kgDOD2dzmg | User | false |
openclaw/openclaw | 4,042,627,469 | I_kwDOQb6kR87w9ZmN | 40,399 | https://github.com/openclaw/openclaw/issues/40399 | https://api.github.com/repos/openclaw/openclaw/issues/40399 | feat: Adaptive Model Routing — run local model first, escalate to cloud on quality failure | ## Summary
Users running local models (e.g. Ollama) to save costs often get insufficient responses on complex tasks and have to manually retry with a cloud model. There's no automated way to get the best of both worlds — cheap local for simple queries, cloud quality when it actually matters.
**Proposal:** Add an opt-... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-09T00:33:26Z | 2026-04-30T04:49:38Z | 2026-04-25T05:49:57Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | joelnishanth | 140,015,627 | U_kgDOCFh4Cw | User | false |
openclaw/openclaw | 4,042,845,502 | I_kwDOQb6kR87w-O0- | 40,445 | https://github.com/openclaw/openclaw/issues/40445 | https://api.github.com/repos/openclaw/openclaw/issues/40445 | feat: passive execution health monitoring for agent death spirals | ## Problem
Agents can enter execution death spirals where they produce artifacts (planning files, checklists, markdown documents) instead of executing real work. This happens when:
1. The agent hits a persistent blocker (auth failure, rate limit, tool unavailability)
2. A tight SLA watchdog or heartbeat forces the ag... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-09T01:49:44Z | 2026-04-30T04:49:37Z | 2026-04-25T05:50:04Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dsantoreis | 220,753,637 | U_kgDODShu5Q | User | false |
openclaw/openclaw | 4,043,060,407 | I_kwDOQb6kR87w_DS3 | 40,505 | https://github.com/openclaw/openclaw/issues/40505 | https://api.github.com/repos/openclaw/openclaw/issues/40505 | [Bug]: the structure of feishu channel config in openclaw.json differs from the demonstration on the website. | ### Bug type
Regression (worked before, now fails)
### Summary
[Bug]: the structure of feishu channel config in openclaw.json differs from the demonstration on the website.
in config:
"channels": {
"feishu": {
"enabled": true,
"appId": "cli",
"appSecret": "",
"connectionMode": "websocket"... | closed | completed | true | 2 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-09T03:09:23Z | 2026-04-30T04:49:36Z | 2026-04-25T05:50:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Zhitao-JIn | 237,718,813 | U_kgDODitNHQ | User | false |
openclaw/openclaw | 4,043,170,358 | I_kwDOQb6kR87w_eI2 | 40,553 | https://github.com/openclaw/openclaw/issues/40553 | https://api.github.com/repos/openclaw/openclaw/issues/40553 | Feature: Auto-merge workspace NOTES.md into bundled skills at load time | ## Summary
When users accumulate tool-specific operational lessons (e.g., gotchas about CLI behavior, auth patterns, query strategies), those lessons need to persist across sessions and survive skill updates. Currently this requires overriding the entire bundled skill with a workspace copy + sync script.
## Proposed ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-09T03:51:06Z | 2026-04-30T04:49:35Z | 2026-04-25T05:50:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | norfolkailabs | 263,759,820 | U_kgDOD7inzA | User | false |
openclaw/openclaw | 4,327,235,411 | I_kwDOQb6kR88AAAABAexfUw | 71,412 | https://github.com/openclaw/openclaw/issues/71412 | https://api.github.com/repos/openclaw/openclaw/issues/71412 | stopChannel abort-timeout leaves zombie task in store, preventing health-monitor restart | ## Summary
When `stopChannel` exceeds `CHANNEL_STOP_ABORT_TIMEOUT_MS` (5 s), the channel runtime is left in a state where `startChannelInternal` silently no-ops and the health-monitor believes the restart succeeded. The channel appears `running: true, connected: true` in the runtime snapshot but no polling loop is act... | open | null | false | 4 | [] | [] | 2026-04-25T05:50:24Z | 2026-05-03T13:21:30Z | null | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | fl8-dev | 44,420,500 | MDQ6VXNlcjQ0NDIwNTAw | User | false |
openclaw/openclaw | 4,043,215,493 | I_kwDOQb6kR87w_pKF | 40,562 | https://github.com/openclaw/openclaw/issues/40562 | https://api.github.com/repos/openclaw/openclaw/issues/40562 | Feature: Cross-App Control — interact with native apps (macOS/iOS) | ## Summary
Agents currently cannot interact with native apps on the host machine or paired nodes. For example, creating a note in a local note-taking app (Apple Notes, Pencil, Obsidian, etc.) from a Telegram conversation is not possible.
## Use Case
User sends screenshots to an agent in Telegram → agent extracts con... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-09T04:05:38Z | 2026-04-30T04:49:35Z | 2026-04-25T05:50:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | howardpen9 | 36,764,495 | MDQ6VXNlcjM2NzY0NDk1 | User | false |
openclaw/openclaw | 4,043,556,349 | I_kwDOQb6kR87xA8X9 | 40,668 | https://github.com/openclaw/openclaw/issues/40668 | https://api.github.com/repos/openclaw/openclaw/issues/40668 | [Bug]: Tools returning MEDIA: paths lack auto-delivery hint, causing agent to misreport media delivery | ## Summary
Tools that return `MEDIA:` paths in their result (browser, nodes, canvas) lack the auto-delivery documentation that the TTS tool already has. This causes the agent to incorrectly claim media was not sent, or to re-send it explicitly via the message tool.
## Steps to reproduce
1. Use the `browser` tool wit... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-09T06:07:10Z | 2026-04-30T04:49:34Z | 2026-04-25T05:50:30Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | carrotRakko | 24,588,751 | MDQ6VXNlcjI0NTg4NzUx | User | false |
openclaw/openclaw | 4,043,584,835 | I_kwDOQb6kR87xBDVD | 40,674 | https://github.com/openclaw/openclaw/issues/40674 | https://api.github.com/repos/openclaw/openclaw/issues/40674 | [Bug]: ACP Claude Code do not support /add-dir command | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
I use OpenClaw 2026.3.7 in Discord. I bind a channel with claude code by ACP.
My website has a frontend dir and a backend dir. The workdir is backend dir.
I say "/add-dir ../frontend" and want claude code to add my frontend dir to the wor... | closed | completed | true | 4 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-09T06:15:12Z | 2026-04-30T04:49:33Z | 2026-04-25T05:50:37Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | qkxie | 134,578,340 | U_kgDOCAWApA | User | false |
openclaw/openclaw | 4,043,667,108 | I_kwDOQb6kR87xBXak | 40,689 | https://github.com/openclaw/openclaw/issues/40689 | https://api.github.com/repos/openclaw/openclaw/issues/40689 | [Feature Request] Add tool:called or tool:invoke event to Hooks system | ## Feature Request: Add tool:called / tool:invoke event to Hooks
### Problem
Currently, OpenClaw Hooks only support `command` events (like `/new`, `/reset`, `/stop`). There's no way to automatically track when Skills or tools are invoked.
### Use Case
I want to automatically log and analyze:
- Which Skills are being ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-09T06:37:19Z | 2026-04-30T04:49:32Z | 2026-04-25T05:50:43Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | a158070653-hash | 263,277,539 | U_kgDOD7FL4w | User | false |
openclaw/openclaw | 4,043,681,142 | I_kwDOQb6kR87xBa12 | 40,695 | https://github.com/openclaw/openclaw/issues/40695 | https://api.github.com/repos/openclaw/openclaw/issues/40695 | [Bug]: Dashboard return "Not Found" | ### Bug type
Regression (worked before, now fails)
### Summary
After upgrading to 2026.3.7, dashboard return 'Not Found'
### Steps to reproduce
1. upgrade to 2026.3.7
2. open dashboard in browser
### Expected behavior
open dashboard in browser correctly
### Actual behavior
dashboard return 'Not Found'
### Ope... | closed | completed | true | 7 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-09T06:41:38Z | 2026-04-30T04:49:31Z | 2026-04-25T05:50:50Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | deadflyy | 11,189,487 | MDQ6VXNlcjExMTg5NDg3 | User | false |
openclaw/openclaw | 4,043,766,572 | I_kwDOQb6kR87xBvss | 40,713 | https://github.com/openclaw/openclaw/issues/40713 | https://api.github.com/repos/openclaw/openclaw/issues/40713 | Discord: forwarded messages are silently dropped on inbound | ## What's happening
When someone forwards a Discord message from one channel to another, the forwarded content is silently stripped from the inbound message. The agent receives an empty message — no text, no context, nothing to work with.
## Why it matters
A natural use case for multi-agent setups is forwarding mess... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-09T07:04:26Z | 2026-04-30T04:49:30Z | 2026-04-25T05:50:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | tedzhao226 | 12,790,328 | MDQ6VXNlcjEyNzkwMzI4 | User | false |
openclaw/openclaw | 4,043,826,529 | I_kwDOQb6kR87xB-Vh | 40,725 | https://github.com/openclaw/openclaw/issues/40725 | https://api.github.com/repos/openclaw/openclaw/issues/40725 | Two Critical Issues: Stale Process After Upgrade & Unsafe Schema Access in Claude Fallback | # Bug Report: Two Critical Issues in OpenClaw 2026.3.7
## Environment
- **OpenClaw Version**: 2026.3.7
- **Node Version**: v24.13.1
- **Platform**: Linux (WSL2)
- **Installation**: npm global install
---
## Issue 1: Stale Process References After Upgrade
### Symptom
After upgrading OpenClaw from an older version to... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-09T07:18:58Z | 2026-04-30T04:49:29Z | 2026-04-25T05:51:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | chenhaonan-eth | 37,938,049 | MDQ6VXNlcjM3OTM4MDQ5 | User | false |
openclaw/openclaw | 4,044,353,076 | I_kwDOQb6kR87xD-40 | 40,806 | https://github.com/openclaw/openclaw/issues/40806 | https://api.github.com/repos/openclaw/openclaw/issues/40806 | [Bug]: Agent tool calls (exec, cron) are isolated and don't affect real filesystem | ### Bug type
Regression (worked before, now fails)
### Summary
Agent tool calls appear to succeed but don't actually affect the host filesystem. Commands return successful output, but files are not created and cron jobs are not persisted
### Steps to reproduce
Ask agent to write a file: "Write a test file to ~/Des... | closed | completed | true | 8 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-09T09:13:12Z | 2026-04-30T04:49:29Z | 2026-04-25T05:51:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | LCDedc | 83,866,934 | MDQ6VXNlcjgzODY2OTM0 | User | false |
openclaw/openclaw | 4,044,738,509 | I_kwDOQb6kR87xFc_N | 40,862 | https://github.com/openclaw/openclaw/issues/40862 | https://api.github.com/repos/openclaw/openclaw/issues/40862 | [Agent-to-Agent] sessions_send should support receive-acknowledge pattern | ## Problem
Currently, `sessions_send` waits for the target agent to fully process and respond before returning. This causes long latency (5-60+ seconds) before the sender gets any feedback.
## Current Behavior
```
Sender → Message → Receiver
↓ (wait for LLM processing)
↓ (wait... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-09T10:27:10Z | 2026-04-30T04:49:28Z | 2026-04-25T05:51:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | colaboy519 | 97,298,765 | U_kgDOBcypTQ | User | false |
openclaw/openclaw | 4,045,074,918 | I_kwDOQb6kR87xGvHm | 40,934 | https://github.com/openclaw/openclaw/issues/40934 | https://api.github.com/repos/openclaw/openclaw/issues/40934 | SECURITY: Malicious skill 'openclaw-email-bypass' in community skills repo — covert email backdoor | ## Security Report
**Affected repo:** `openclaw/skills`
**Skill path:** `skills/rishikreddyl/openclaw-email-bypass`
**Commit:** `458db28ca3cd0fcb5e75120f45ff431866abdff4`
**Merged by:** `clawdhub[bot]` on 2026-02-07
### What it does
This skill is a covert email relay backdoor. It instructs the agent to:
1. Deploy a... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-09T11:37:51Z | 2026-04-30T04:49:27Z | 2026-04-25T05:51:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | pepelaskov | 13,340,443 | MDQ6VXNlcjEzMzQwNDQz | User | false |
openclaw/openclaw | 4,114,878,182 | I_kwDOQb6kR871RA7m | 52,097 | https://github.com/openclaw/openclaw/issues/52097 | https://api.github.com/repos/openclaw/openclaw/issues/52097 | Cron job runs repeatedly after manual trigger via cron run |
After manually triggering a cron job using `cron run` command, the job starts repeatedly outside of its schedule.
## Steps to reproduce
1. Create a cron job with schedule `0 16 * * *` (once daily at 16:00 UTC)
2. Manually trigger the job using `cron run` command
3. After manual execution, the job continues to run rep... | closed | completed | true | 4 | [] | [] | 2026-03-22T05:18:20Z | 2026-04-30T04:49:26Z | 2026-04-25T05:51:30Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | web3blind | 264,741,654 | U_kgDOD8ejFg | User | false |
openclaw/openclaw | 4,115,057,502 | I_kwDOQb6kR871Rste | 52,108 | https://github.com/openclaw/openclaw/issues/52108 | https://api.github.com/repos/openclaw/openclaw/issues/52108 | [Feature] Hook: tool:before-execute event for exec/tool pre-execution interception | ### Summary
Add a hook event that fires before AI executes a tool or exec command, enabling pre-execution approval/interception
### Problem to solve
AI agents can autonomously decide to execute shell commands or tools (via exec tool). Currently there is no mechanism to intercept this decision point. The existing hoo... | closed | completed | true | 2 | [
"enhancement"
] | [] | 2026-03-22T06:06:16Z | 2026-04-30T04:49:25Z | 2026-04-25T05:51:36Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Lrs-GreenBottle | 253,340,193 | U_kgDODxmqIQ | User | false |
openclaw/openclaw | 4,115,090,122 | I_kwDOQb6kR871R0rK | 52,112 | https://github.com/openclaw/openclaw/issues/52112 | https://api.github.com/repos/openclaw/openclaw/issues/52112 | Discord: Thread context lost on gateway restart — no ThreadHistoryBody parity with Slack | ## Summary
When the OpenClaw gateway restarts, Discord thread sessions lose all prior conversation context. Each restart creates a new session file for the same thread, with only the ThreadStarterBody (the very first message) injected — not the recent conversation history. This causes the LLM to lose track of ongoing ... | closed | not_planned | true | 4 | [] | [] | 2026-03-22T06:17:30Z | 2026-04-30T04:49:25Z | 2026-04-25T05:51:43Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dbvrl | 209,000,756 | U_kgDODHUZNA | User | false |
openclaw/openclaw | 4,115,124,934 | I_kwDOQb6kR871R9LG | 52,116 | https://github.com/openclaw/openclaw/issues/52116 | https://api.github.com/repos/openclaw/openclaw/issues/52116 | Telegram polling client gets permanently stuck after transient network failure | ## Description
The gateway's Telegram client enters an unrecoverable state after a transient network hiccup. Once it fails, it retries `sendChatAction` in a tight loop (~every 3 seconds) indefinitely, even across gateway restarts.
## Evidence
- `curl` can successfully call the Telegram Bot API from the same machine ... | closed | completed | true | 7 | [] | [] | 2026-03-22T06:36:16Z | 2026-04-30T04:49:24Z | 2026-04-25T05:51:49Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | rudniai | 263,499,678 | U_kgDOD7Svng | User | false |
openclaw/openclaw | 4,115,308,266 | I_kwDOQb6kR871Sp7q | 52,145 | https://github.com/openclaw/openclaw/issues/52145 | https://api.github.com/repos/openclaw/openclaw/issues/52145 | [Bug]: openviking plugin can leak agent context via shared mutable client agentId | ## Problem
The OpenClaw `openviking` context-engine plugin appears to use a shared mutable client whose `agentId` is reassigned per session/request.
This creates a concurrency risk: when multiple sessions or agents hit the shared client at the same time, one request can overwrite the `agentId` intended for another.
... | closed | not_planned | true | 4 | [] | [] | 2026-03-22T08:15:44Z | 2026-04-30T04:49:23Z | 2026-04-25T05:51:56Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ningfeemic-dev | 239,259,264 | U_kgDODkLOgA | User | false |
openclaw/openclaw | 4,115,448,593 | I_kwDOQb6kR871TMMR | 52,167 | https://github.com/openclaw/openclaw/issues/52167 | https://api.github.com/repos/openclaw/openclaw/issues/52167 | [Bug]: WhatsApp media sends fail on old-style group JIDs — message tool bypassed on new-style JIDs | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
WhatsApp media sends via the `message` tool fail with `No active WhatsApp Web listener (account: default)` on **old-style group JIDs** (format `XXXXXXXXXXX-XXXXXXXXXX@g.us`), while text auto-replies work fine on the same groups, and both te... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T09:23:31Z | 2026-04-30T04:49:21Z | 2026-04-25T05:52:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ccxdomo | 194,931,446 | U_kgDOC55q9g | User | false |
openclaw/openclaw | 4,115,454,433 | I_kwDOQb6kR871TNnh | 52,169 | https://github.com/openclaw/openclaw/issues/52169 | https://api.github.com/repos/openclaw/openclaw/issues/52169 | [Bug] Web UI 语音输入无法使用 - Content Security Policy 阻止 | ## 问题描述
在 OpenClaw Web UI (http://127.0.0.1:18789) 中，自带的麦克风按钮无法正常工作。
## 环境
- OpenClaw 版本: 2026.3.13
- 浏览器: Chrome (Edge)
- 操作系统: Windows 11 / WSL2
## 复现步骤
1. 打开 OpenClaw Web UI
2. 点击左下角的麦克风按钮
3. 按住按钮说话
## 实际结果
浏览器控制台报错：
错误详情：
- 问题出在 被 CSP 阻止
- 这导致语音输入功能完全无法使用
- 语音播放也被阻止
## 期望结果
用户应该能够：
1. 通过麦克风按钮进行语音输入
2. ... | closed | completed | true | 2 | [] | [] | 2026-03-22T09:26:03Z | 2026-04-30T04:49:21Z | 2026-04-25T05:52:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | acientemple | 43,695,197 | MDQ6VXNlcjQzNjk1MTk3 | User | false |
openclaw/openclaw | 4,115,497,338 | I_kwDOQb6kR871TYF6 | 52,177 | https://github.com/openclaw/openclaw/issues/52177 | https://api.github.com/repos/openclaw/openclaw/issues/52177 | web_search tool fails with 'fetch failed' despite valid Brave API key | ## Bug Report
**Environment:**
- OpenClaw version: 2026.3.13
- OS: macOS (Darwin 25.2.0)
- Node: v24.4.1
**Issue:**
The `web_search` tool consistently fails with `fetch failed` error, even though:
- Brave API key is valid and configured in `openclaw.json`
- Direct curl calls to Brave API work perfectly
- `OPENCLAW_WE... | closed | completed | true | 1 | [] | [] | 2026-03-22T09:43:32Z | 2026-04-30T04:49:19Z | 2026-04-25T05:52:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | qiankemeng | 68,374,326 | MDQ6VXNlcjY4Mzc0MzI2 | User | false |
openclaw/openclaw | 4,115,504,644 | I_kwDOQb6kR871TZ4E | 52,180 | https://github.com/openclaw/openclaw/issues/52180 | https://api.github.com/repos/openclaw/openclaw/issues/52180 | [Bug]: Podman Windows WSL permission denied | ### Bug type
Regression (worked before, now fails)
### Summary
Loading image into openclaw Podman store...
Error: faccessat /var/tmp/openclaw-podman-image..../openclaw-image.tar: permission denied
### Steps to reproduce
Running ./scripts/podman/setup.sh
on Win. WSL with podman
Latest github
### Expected behavior
... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-22T09:47:39Z | 2026-04-30T04:49:18Z | 2026-04-25T05:52:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | theoren | 2,540,177 | MDQ6VXNlcjI1NDAxNzc= | User | false |
openclaw/openclaw | 4,115,608,086 | I_kwDOQb6kR871TzIW | 52,189 | https://github.com/openclaw/openclaw/issues/52189 | https://api.github.com/repos/openclaw/openclaw/issues/52189 | [Bug]: session_status reports default model instead of channels.modelByChannel effective model | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using channels.modelByChannel for IRC, the actual routed model is correct, but session_status reports the default agent model instead of the effective channel-selected model
Version
Observed on: OpenClaw 2026.3.13
Config
"channels": ... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T10:27:48Z | 2026-04-30T04:49:17Z | 2026-04-25T05:52:42Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | khanmore | 30,566,690 | MDQ6VXNlcjMwNTY2Njkw | User | false |
openclaw/openclaw | 4,115,649,595 | I_kwDOQb6kR871T9Q7 | 52,197 | https://github.com/openclaw/openclaw/issues/52197 | https://api.github.com/repos/openclaw/openclaw/issues/52197 | Platform Setup Review: Compare OpenClaw GitHub Integration vs gh CLI Capabilities | ## Objective
Review current OpenClaw GitHub integration and workflows, compare to native gh CLI capabilities, and identify gaps/opportunities for follow-up.
---
## Current State
### OpenClaw GitHub Integration
**gh-issues Skill:**
- Fetch issues via GitHub REST API (not gh CLI)
- Spawn sub-agents to implement fixe... | closed | not_planned | true | 2 | [] | [] | 2026-03-22T10:44:17Z | 2026-04-30T04:49:15Z | 2026-04-25T05:52:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MrMoniker | 20,151,755 | MDQ6VXNlcjIwMTUxNzU1 | User | false |
openclaw/openclaw | 4,115,661,601 | I_kwDOQb6kR871UAMh | 52,199 | https://github.com/openclaw/openclaw/issues/52199 | https://api.github.com/repos/openclaw/openclaw/issues/52199 | Memory search CLI vs agent internal behavior inconsistency + poor error messages | ## Summary
Multiple UX and reliability issues discovered during memory search configuration with a custom embedding provider (OpenAI-compatible proxy). Documented from a real debugging session on 2026-03-22.
---
## Issue 1: CLI `openclaw memory search` and agent internal `memory_search` take different code paths wit... | closed | completed | true | 3 | [] | [] | 2026-03-22T10:50:01Z | 2026-04-30T04:49:14Z | 2026-04-25T05:53:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | iceseaboy | 249,952 | MDQ6VXNlcjI0OTk1Mg== | User | false |
openclaw/openclaw | 4,115,690,828 | I_kwDOQb6kR871UHVM | 52,204 | https://github.com/openclaw/openclaw/issues/52204 | https://api.github.com/repos/openclaw/openclaw/issues/52204 | [Bug] 多渠道会话隔离问题：消息路由到错误的渠道 (session isolation & channel routing bug) | ## Bug Description
当同时启用飞书和微信两个渠道时，OpenClaw 会绕过系统自动路由机制，导致消息发送到错误的渠道。
## Environment
- **OpenClaw version:** 2026.3.13
- **Channels enabled:** feishu, openclaw-weixin
- **Primary agent:** main
## Reproduction Steps
1. 配置飞书渠道（channels.feishu.enabled = true）
2. 配置微信渠道（channels.wechat.enabled = true，通过 openclaw-weixi... | closed | completed | true | 2 | [] | [] | 2026-03-22T11:04:04Z | 2026-04-30T04:49:14Z | 2026-04-25T05:53:08Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Designer-Awei | 200,909,454 | U_kgDOC_mijg | User | false |
openclaw/openclaw | 4,115,724,405 | I_kwDOQb6kR871UPh1 | 52,206 | https://github.com/openclaw/openclaw/issues/52206 | https://api.github.com/repos/openclaw/openclaw/issues/52206 | [Bug]: OpenClaw v2026.3.13 still sends Ollama an oversized context window (~262144 / 265k), ignoring configured lower limits | ### Bug type
Regression (worked before, now fails)
### Summary
### Summary
OpenClaw `v2026.3.13` still appears to send an extremely large context window to Ollama (around `262144` / `265k`), even when a much lower context size is configured.
This looks very similar to the previously reported issue about OpenClaw f... | closed | completed | true | 4 | [
"bug",
"regression"
] | [] | 2026-03-22T11:18:40Z | 2026-04-30T04:49:13Z | 2026-04-25T05:53:14Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | LokiCode404 | 1,407,829 | MDQ6VXNlcjE0MDc4Mjk= | User | false |
openclaw/openclaw | 4,213,240,445 | I_kwDOQb6kR877IPJ9 | 62,059 | https://github.com/openclaw/openclaw/issues/62059 | https://api.github.com/repos/openclaw/openclaw/issues/62059 | [Feishu] Card streaming creates duplicate card on every response (counts.final=2) | ### Environment
- OpenClaw version: `2026.4.5 (3e72c03)`
- Feishu plugin version: `2026.3.8`
- Channel: Feishu (card mode, streaming enabled)
- Node: v22.22.1, Linux arm64
### Description
Every response dispatch creates **two streaming card messages** in Feishu DMs. The first card receives the full streamed content, a... | closed | completed | true | 3 | [
"close:already-fixed",
"dedupe:child"
] | [
"vincentkoc"
] | 2026-04-06T18:07:19Z | 2026-04-30T04:47:36Z | 2026-04-25T05:52:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | longdoubled7 | 25,959,199 | MDQ6VXNlcjI1OTU5MTk5 | User | false |
openclaw/openclaw | 4,115,732,768 | I_kwDOQb6kR871URkg | 52,208 | https://github.com/openclaw/openclaw/issues/52208 | https://api.github.com/repos/openclaw/openclaw/issues/52208 | [Bug]: macOS LaunchAgent can be removed and left not loaded after failed `openclaw gateway start` | ### Bug type
Regression (worked before, now fails)
### Summary
## Environment
- OpenClaw version: `2026.3.13 (61d171a)`
- OS: macOS
- Service manager: launchd / LaunchAgent
- Service plist: `~/Library/LaunchAgents/ai.openclaw.gateway.plist`
## Summary
I hit a failure mode where the OpenClaw Gateway stopped and did... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-22T11:22:58Z | 2026-04-30T04:49:11Z | 2026-04-25T05:53:21Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | alex-quan-001 | 77,444,503 | MDQ6VXNlcjc3NDQ0NTAz | User | false |
openclaw/openclaw | 4,086,004,313 | I_kwDOQb6kR87zi3pZ | 48,699 | https://github.com/openclaw/openclaw/issues/48699 | https://api.github.com/repos/openclaw/openclaw/issues/48699 | [Feishu] Streaming card fallback sends duplicate messages when streaming fails | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
duplicates on feishu card messages
### Steps to reproduce
## Bug Description
When `streaming: true` and `renderMode: "card"` are enabled in Feishu channel config, the bot sends **duplicate messages** - one failed streaming attempt + one ... | closed | completed | true | 3 | [
"bug",
"close:already-fixed",
"dedupe:child",
"bug:behavior"
] | [
"vincentkoc"
] | 2026-03-17T03:34:45Z | 2026-04-30T04:49:12Z | 2026-04-25T05:52:39Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | bearhunt110 | 38,199,735 | MDQ6VXNlcjM4MTk5NzM1 | User | false |
openclaw/openclaw | 4,075,792,089 | I_kwDOQb6kR87y76bZ | 46,135 | https://github.com/openclaw/openclaw/issues/46135 | https://api.github.com/repos/openclaw/openclaw/issues/46135 | Feishu Channel: Duplicate streaming cards due to state management race condition in reply-dispatcher.ts | ## Bug Description
Multiple streaming cards are created for a single conversational turn in the Feishu channel, resulting in duplicate messages being sent to users.
## Root Cause
The issue lies in the state management of `streaming` and `streamingStartPromise` within `/home/jonathan/.npm-global/lib/node_modules/open... | closed | completed | true | 2 | [
"close:already-fixed",
"dedupe:child"
] | [
"vincentkoc"
] | 2026-03-14T13:09:29Z | 2026-04-30T04:49:10Z | 2026-04-25T05:53:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jojojonathan | 37,664,587 | MDQ6VXNlcjM3NjY0NTg3 | User | false |
openclaw/openclaw | 4,115,758,581 | I_kwDOQb6kR871UX31 | 52,211 | https://github.com/openclaw/openclaw/issues/52211 | https://api.github.com/repos/openclaw/openclaw/issues/52211 | [Feature Request] 企微插件支持显示 Reasoning/Thinking 思考过程 | ## 功能请求
**通道**：企业微信（wecom-openclaw-plugin v1.0.11）
**问题描述**：
Telegram 和飞书插件支持通过 \`/reasoning on\` 或 \`/reasoning stream\` 显示 AI 推理过程，但企业微信插件不支持显示思考过程内容。
企微插件的配置中没有任何 thinking/reasoning 相关的选项，用户无法看到 AI 的推理思考过程。
**期望行为**：
希望企微插件也能像 Telegram 和飞书一样，支持显示 AI 的推理思考过程（reasoning/thinking content），让用户能够：
1. 看到 AI 的思考过程（类似 Te... | closed | not_planned | true | 1 | [] | [] | 2026-03-22T11:35:49Z | 2026-04-30T04:49:10Z | 2026-04-25T05:53:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Lrs-GreenBottle | 253,340,193 | U_kgDODxmqIQ | User | false |
openclaw/openclaw | 4,163,940,712 | I_kwDOQb6kR874MLFo | 56,981 | https://github.com/openclaw/openclaw/issues/56981 | https://api.github.com/repos/openclaw/openclaw/issues/56981 | [Feishu] Streaming card fails with HTTP 400, causing ~2-3s delay per message | ## Problem
Every message sent via Feishu channel triggers an HTTP 400 error when attempting to create a streaming card:
```
feishu: streaming start failed: Error: Create card request failed with HTTP 400
```
This causes approximately 2-3 seconds additional delay per message (failure wait + retry).
## Environment
-... | closed | completed | true | 2 | [] | [
"vincentkoc"
] | 2026-03-29T11:46:39Z | 2026-04-30T04:45:41Z | 2026-04-25T06:41:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JinnanDuan | 41,154,709 | MDQ6VXNlcjQxMTU0NzA5 | User | false |
openclaw/openclaw | 4,115,771,693 | I_kwDOQb6kR871UbEt | 52,214 | https://github.com/openclaw/openclaw/issues/52214 | https://api.github.com/repos/openclaw/openclaw/issues/52214 | [Feature]: My project encountered some issues during the iteration of the 2.0 version. | ### Summary
Github：https://github.com/openpup/openpup
### Problem to solve
Rt
### Proposed solution
Rt
### Alternatives considered
Rt
### Impact
Rt
### Evidence/examples
_No response_
### Additional information
_No response_ | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-22T11:41:48Z | 2026-04-30T04:49:09Z | 2026-04-25T05:53:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | BenalexCheung | 25,890,736 | MDQ6VXNlcjI1ODkwNzM2 | User | false |
openclaw/openclaw | 4,115,795,043 | I_kwDOQb6kR871Ugxj | 52,220 | https://github.com/openclaw/openclaw/issues/52220 | https://api.github.com/repos/openclaw/openclaw/issues/52220 | Restart sentinel notification lost: sentinel consumed before Telegram channel is ready | ## Summary
After a `config.patch` restart, the restart sentinel notification is silently lost because the sentinel is consumed ~750ms after startup, but the Telegram bot provider doesn't start until ~3s after startup.
## Steps to Reproduce
1. Run `config.patch` with a `note` parameter
2. Gateway restarts via SIGUSR1... | closed | completed | true | 2 | [] | [] | 2026-03-22T11:49:32Z | 2026-04-30T04:49:08Z | 2026-04-25T05:53:41Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | yyxiaohui | 235,090,347 | U_kgDODgMxqw | User | false |
openclaw/openclaw | 4,115,829,615 | I_kwDOQb6kR871UpNv | 52,224 | https://github.com/openclaw/openclaw/issues/52224 | https://api.github.com/repos/openclaw/openclaw/issues/52224 | Zombie embedded run blocks heartbeats permanently after provider timeout | ## Summary
When an embedded run times out but the underlying provider promise never settles (e.g., dead connection, hung HTTP stream), the run handle stays in `ACTIVE_EMBEDDED_RUNS` permanently. Since `resolveActiveRunQueueAction` returns `"drop"` for heartbeats when `isActive === true`, all subsequent heartbeat ticks... | closed | completed | true | 2 | [] | [] | 2026-03-22T12:01:49Z | 2026-04-30T04:49:05Z | 2026-04-25T05:53:54Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | habonlaci | 4,699,494 | MDQ6VXNlcjQ2OTk0OTQ= | User | false |
openclaw/openclaw | 4,115,834,025 | I_kwDOQb6kR871UqSp | 52,225 | https://github.com/openclaw/openclaw/issues/52225 | https://api.github.com/repos/openclaw/openclaw/issues/52225 | [Feature]: Gateway relay endpoint — give heartbeat-only agents a public A2A URL | ## Problem
OpenClaw agents run in heartbeat sessions: no public IP, no inbound ports, no persistent server. This means they can't register on Agora (https://the-agora.dev) or any A2A-compatible agent registry, because registration requires a publicly reachable URL that:
1. Responds to `GET /.well-known/agent-card.jso... | closed | not_planned | true | 4 | [] | [] | 2026-03-22T12:03:51Z | 2026-04-28T04:30:29Z | 2026-04-25T05:54:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | archedark-ada | 258,527,876 | U_kgDOD2jShA | User | false |
openclaw/openclaw | 4,115,856,058 | I_kwDOQb6kR871Uvq6 | 52,231 | https://github.com/openclaw/openclaw/issues/52231 | https://api.github.com/repos/openclaw/openclaw/issues/52231 | [Bug]: Embedded run timeout leaves zombie handle blocking heartbeat delivery | ## Bug Summary
When an embedded run times out but the underlying provider promise never settles (e.g., dead HTTP connection, hung stream), the run handle stays in `ACTIVE_EMBEDDED_RUNS` permanently. This silently kills all subsequent heartbeat deliveries for the session.
## Steps to Reproduce
1. Configure a heartbea... | closed | completed | true | 3 | [] | [] | 2026-03-22T12:15:43Z | 2026-04-30T04:49:05Z | 2026-04-25T05:54:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ai-nurmamat | 130,913,840 | U_kgDOB82WMA | User | false |
openclaw/openclaw | 4,115,861,488 | I_kwDOQb6kR871Uw_w | 52,233 | https://github.com/openclaw/openclaw/issues/52233 | https://api.github.com/repos/openclaw/openclaw/issues/52233 | [Bug]: The UI fails to correctly update the provider prefix when switching models across different providers, resulting in invalid model paths. | ### Bug type
Regression (worked before, now fails)
### Summary
The UI fails to correctly update the provider prefix when switching models across different providers, resulting in invalid model paths.
### Steps to reproduce
The UI fails to correctly update the provider prefix when switching models across different ... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-22T12:18:51Z | 2026-04-30T04:49:04Z | 2026-04-25T05:54:14Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | yzullf | 264,913,328 | U_kgDOD8pBsA | User | false |
openclaw/openclaw | 4,115,917,633 | I_kwDOQb6kR871U-tB | 52,243 | https://github.com/openclaw/openclaw/issues/52243 | https://api.github.com/repos/openclaw/openclaw/issues/52243 | [Feature]: Expose inbound image attachments as local file paths for agent tool use | ## Feature Description
When a user sends an image to an agent via a messaging channel (Signal, Telegram, WebChat, etc.), the agent can **receive and understand** the image content through the vision model. However, the agent has **no way to access the image as a local file** for downstream operations such as:
- Uploa... | closed | completed | true | 1 | [] | [] | 2026-03-22T12:39:47Z | 2026-04-30T04:49:03Z | 2026-04-25T05:54:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Antsoldier1974 | 248,194,488 | U_kgDODssluA | User | false |
openclaw/openclaw | 4,115,935,049 | I_kwDOQb6kR871VC9J | 52,248 | https://github.com/openclaw/openclaw/issues/52248 | https://api.github.com/repos/openclaw/openclaw/issues/52248 | sessions_yield leaves control-ui frozen: WebSocket RPC response routes to wrong connection | When a WebSocket client (control-ui) calls `agent.wait` RPC and a parent turn completes, the response is routed to an ephemeral internal connection instead of the original client connection. The browser UI remains frozen/waiting until the user manually refreshes.
## Steps to Reproduce
1. Open control-ui in browser
2.... | closed | not_planned | true | 3 | [] | [] | 2026-03-22T12:44:36Z | 2026-04-30T04:49:02Z | 2026-04-25T05:54:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dapzthelegend | 47,679,048 | MDQ6VXNlcjQ3Njc5MDQ4 | User | false |
openclaw/openclaw | 4,327,249,242 | I_kwDOQb6kR88AAAABAeyVWg | 71,414 | https://github.com/openclaw/openclaw/issues/71414 | https://api.github.com/repos/openclaw/openclaw/issues/71414 | Heartbeat duration `every` >24.85d overflows Node setTimeout, crashes gateway with no auto-respawn | ## Summary
Setting `agents.defaults.heartbeat.every` to a duration greater than ~24.85 days (Node.js's signed-32-bit `setTimeout` cap of 2,147,483,647 ms) causes the heartbeat scheduler to fire in a tight loop and eventually crashes the gateway with `OpenClaw exited with code 1`. The container wrapper does not auto-re... | closed | completed | false | 0 | [] | [] | 2026-04-25T05:54:35Z | 2026-04-25T09:03:45Z | 2026-04-25T09:03:44Z | NONE | null | 20260425T120014Z | 2026-04-25T12:00:14Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,327,249,475 | I_kwDOQb6kR88AAAABAeyWQw | 71,415 | https://github.com/openclaw/openclaw/issues/71415 | https://api.github.com/repos/openclaw/openclaw/issues/71415 | CLI `--session-id <new-uuid>` silently remapped to existing agent:<id>:main session | ## Summary
`openclaw agent --agent <id> --session-id <new-uuid>` is silently ignored when an active `agent:<id>:main` session already exists in `agents/<id>/sessions/sessions.json`. The new UUID is accepted on the command line but the run is routed through the existing session, leaking stale conversation history into ... | closed | completed | true | 1 | [] | [] | 2026-04-25T05:54:36Z | 2026-04-28T04:46:52Z | 2026-04-25T19:32:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,327,249,504 | I_kwDOQb6kR88AAAABAeyWYA | 71,416 | https://github.com/openclaw/openclaw/issues/71416 | https://api.github.com/repos/openclaw/openclaw/issues/71416 | CLI silently falls back to embedded mode when gateway is unreachable | ## Summary
When the gateway is unreachable, `openclaw agent` silently falls back to embedded mode and writes a single line to `stderr` that's easy to miss. There's no banner on `stdout`, no exit-code change, and the JSON `result` looks superficially like a normal gateway response. Operators who don't `tail .err` will ... | closed | completed | true | 1 | [] | [] | 2026-04-25T05:54:37Z | 2026-04-30T04:34:12Z | 2026-04-27T12:14:12Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,327,249,527 | I_kwDOQb6kR88AAAABAeyWdw | 71,417 | https://github.com/openclaw/openclaw/issues/71417 | https://api.github.com/repos/openclaw/openclaw/issues/71417 | `openclaw agent` defaults --channel to last, silently resumes most recent session | ## Summary
`openclaw agent --agent <id> -m "..."` invoked without `--session-id` defaults `--channel` to `last`, which silently resumes whatever session was most recently active. There's no warning, no banner, and no easy way for an operator running a fresh test to discover this without reading the resulting `agentMet... | open | null | false | 3 | [] | [] | 2026-04-25T05:54:38Z | 2026-05-03T11:49:49Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,327,249,649 | I_kwDOQb6kR88AAAABAeyW8Q | 71,418 | https://github.com/openclaw/openclaw/issues/71418 | https://api.github.com/repos/openclaw/openclaw/issues/71418 | ERR_MODULE_NOT_FOUND for `subagent-registry.runtime.js` on every gateway boot | ## Summary
On every gateway boot, the openclaw-deep-observability plugin's hook registration produces three `[warn] subagent cleanup finalize failed` entries because an internal module file is missing from the published `dist/`:
```
[warn] subagent cleanup finalize failed (90a1fefc-8de5-430f-9a7f-2b84beb8628c): Error... | closed | completed | true | 1 | [] | [] | 2026-04-25T05:54:39Z | 2026-04-28T04:46:51Z | 2026-04-25T19:32:51Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,327,249,879 | I_kwDOQb6kR88AAAABAeyX1w | 71,419 | https://github.com/openclaw/openclaw/issues/71419 | https://api.github.com/repos/openclaw/openclaw/issues/71419 | Plugin ownership policy disagreement between CLI parser and gateway loader | ## Summary
The CLI's pre-flight config parser and the gateway's plugin loader disagree on plugin-directory ownership policy. The CLI emits warnings like:
```
Config warnings:
- plugins: plugin: blocked plugin candidate: suspicious ownership (/data/.openclaw/extensions/mycelium, uid=1000, expected uid=0 or root)
- plu... | closed | not_planned | true | 1 | [] | [] | 2026-04-25T05:54:40Z | 2026-04-28T04:46:50Z | 2026-04-25T19:32:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,116,032,327 | I_kwDOQb6kR871VatH | 52,258 | https://github.com/openclaw/openclaw/issues/52258 | https://api.github.com/repos/openclaw/openclaw/issues/52258 | Feature Request: Support token usage tracking for Bailian Coding Plan | ## Problem
When using Alibaba Cloud Bailian Coding Plan, the API does not return token usage in the response. As a result:
- `/status` always shows `Context: 0/203k (0%)`
- OpenClaw cannot track actual context usage
- Users cannot see how much context window is being consumed
## Root Cause
Coding Plan uses a fixed ... | closed | completed | true | 3 | [] | [] | 2026-03-22T13:13:15Z | 2026-04-30T04:49:01Z | 2026-04-25T05:54:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Dmoist | 59,716,623 | MDQ6VXNlcjU5NzE2NjIz | User | false |
openclaw/openclaw | 4,116,109,879 | I_kwDOQb6kR871Vto3 | 52,269 | https://github.com/openclaw/openclaw/issues/52269 | https://api.github.com/repos/openclaw/openclaw/issues/52269 | [Bug]: 最新的源码在安装飞书消息渠道时报错 | ### Bug type
Regression (worked before, now fails)
### Summary
Error: Config validation failed: plugins: plugin: extension entry escapes package directory: ./index.js
<img width="1600" height="716" alt="Image" src="https://github.com/user-attachments/assets/58262e3d-c573-4394-a109-72daa52d0cc5" />
### Steps to rep... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-22T13:48:29Z | 2026-04-30T04:48:59Z | 2026-04-25T05:54:57Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | flyhighergit | 11,091,133 | MDQ6VXNlcjExMDkxMTMz | User | false |
openclaw/openclaw | 4,116,140,592 | I_kwDOQb6kR871V1Iw | 52,280 | https://github.com/openclaw/openclaw/issues/52280 | https://api.github.com/repos/openclaw/openclaw/issues/52280 | gateway probe times out on loopback; deep audit reports missing scope operator.read (device-identity GatewayClient works) | ### Summary
On a loopback-only local gateway, `openclaw gateway probe` consistently times out / never reaches an RPC-ok state. This cascades into:
- `openclaw status --all` showing: `Gateway ... unreachable (missing scope: operator.read)`
- `openclaw security audit --deep` warning: `gateway.probe_failed (deep) missing... | closed | completed | true | 2 | [] | [] | 2026-03-22T14:03:01Z | 2026-04-30T04:48:57Z | 2026-04-25T05:55:11Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | assitantforjess | 265,343,676 | U_kgDOD9DSvA | User | false |
openclaw/openclaw | 4,116,181,664 | I_kwDOQb6kR871V_Kg | 52,285 | https://github.com/openclaw/openclaw/issues/52285 | https://api.github.com/repos/openclaw/openclaw/issues/52285 | [Bug]: Heartbeat messages will not be sent to Telegram | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Heartbeat is running normally, but Telegram message push is extremely unstable. On very rare occasions, it triggers a completely normal message push; most of the time, it just sends "HEARTBEAT_OK"; and in a few cases, it does not send any m... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T14:21:25Z | 2026-04-30T04:48:56Z | 2026-04-25T05:55:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | hexoctal-sur | 13,842,432 | MDQ6VXNlcjEzODQyNDMy | User | false |
openclaw/openclaw | 4,116,190,297 | I_kwDOQb6kR871WBRZ | 52,289 | https://github.com/openclaw/openclaw/issues/52289 | https://api.github.com/repos/openclaw/openclaw/issues/52289 | feat: Auto-cleanup stale session lock files on gateway startup | ## Problem
Session lock files (`.lock` in `sessions/`) become stale after gateway restarts. The lock file stores `{ "pid": <old_pid>, "createdAt": <timestamp> }`, but after SIGUSR1 restart the old process dies and the lock file remains with a dead PID reference.
This causes:
- `openclaw doctor` reports "Found N sessi... | closed | completed | true | 2 | [] | [] | 2026-03-22T14:25:29Z | 2026-04-30T04:48:56Z | 2026-04-25T05:55:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | 120106835 | 57,889,251 | MDQ6VXNlcjU3ODg5MjUx | User | false |
openclaw/openclaw | 4,116,216,591 | I_kwDOQb6kR871WHsP | 52,294 | https://github.com/openclaw/openclaw/issues/52294 | https://api.github.com/repos/openclaw/openclaw/issues/52294 | Bug: 子代理会话消息被错误注入到主会话 | # Bug Report: 子代理会话消息被错误注入到主会话
## 基本信息
- **Reporter**: OpenClaw AI Agent (不是人类用户)
- **Date**: 2026-03-22
- **Environment**: macOS, OpenClaw 最新版
---
## 问题描述
使用 `openclaw agent --message` 命令启动子代理后，子代理的消息被错误地注入到主会话历史中，表现为"用户消息"角色。
## 复现步骤
1. 执行 `openclaw agent --message "xxx"` 命令启动子代理
2. 等待命令完成
3. 查看主会话历史 (`~/.open... | closed | completed | true | 4 | [] | [] | 2026-03-22T14:37:41Z | 2026-04-30T04:48:53Z | 2026-04-25T05:55:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | JAANG131 | 203,095,831 | U_kgDODBr_Fw | User | false |
openclaw/openclaw | 4,116,232,192 | I_kwDOQb6kR871WLgA | 52,298 | https://github.com/openclaw/openclaw/issues/52298 | https://api.github.com/repos/openclaw/openclaw/issues/52298 | [Bug]: | ### Bug type
Regression (worked before, now fails)
### Summary
Bug: Subagent completion event leaks into chat, bypasses NO_REPLY
What happens: When an isolated subagent task completes and replies NO_REPLY, the completion event + internal metadata still surfaces in the chat as a user-visible message.
Why it happens... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-22T14:44:20Z | 2026-04-30T04:48:52Z | 2026-04-25T05:55:50Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MNight75 | 36,059,736 | MDQ6VXNlcjM2MDU5NzM2 | User | false |
openclaw/openclaw | 4,116,244,396 | I_kwDOQb6kR871WOes | 52,300 | https://github.com/openclaw/openclaw/issues/52300 | https://api.github.com/repos/openclaw/openclaw/issues/52300 | Feature: /restart command should provide feedback after completion | ## Summary
Currently, when using the `/restart` command in channels (Telegram, Discord, etc.), the command triggers a gateway restart but does not provide any feedback to the user about whether the restart succeeded or failed.
This creates uncertainty for users - they don\'t know if:
- The restart was successful
- Th... | closed | completed | true | 1 | [] | [] | 2026-03-22T14:48:58Z | 2026-04-30T04:48:52Z | 2026-04-25T05:55:57Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cloudpaw5216 | 267,859,441 | U_kgDOD_c18Q | User | false |
openclaw/openclaw | 4,116,259,546 | I_kwDOQb6kR871WSLa | 52,306 | https://github.com/openclaw/openclaw/issues/52306 | https://api.github.com/repos/openclaw/openclaw/issues/52306 | Embedded run holds session-write-lock for 30 minutes when Telegram streaming stalls (watchdog timeout: 600s, not configurable) | **Summary:**
When a user sends a Telegram message that triggers an embedded run, if Telegram polling stalls mid-streaming (e.g., due to lane congestion or CPU contention), the embedded run's streaming output pipeline blocks. The embedded run does not receive a cancellation/failure signal — it continues waiting for stre... | closed | completed | true | 4 | [] | [] | 2026-03-22T14:56:44Z | 2026-04-30T04:48:51Z | 2026-04-25T05:56:03Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | fengqve | 533,341 | MDQ6VXNlcjUzMzM0MQ== | User | false |
openclaw/openclaw | 4,116,283,067 | I_kwDOQb6kR871WX67 | 52,311 | https://github.com/openclaw/openclaw/issues/52311 | https://api.github.com/repos/openclaw/openclaw/issues/52311 | [Bug]: openai的模型会显示成 anthropic/gpt-5.4-pro | ### Bug type
Regression (worked before, now fails)
### Summary
现在雷打不动，依然是 anthropic/gpt-5.4-pro 在跑
<img width="824" height="440" alt="Image" src="https://github.com/user-attachments/assets/a85dd1b1-714a-457e-9048-ed76aadca680" />
### Steps to reproduce
opcnclaw3.13
### Expected behavior
opcnclaw3.13
### Actual... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-22T15:06:47Z | 2026-04-30T04:48:50Z | 2026-04-25T05:56:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | liu366811-hue | 244,927,607 | U_kgDODplMdw | User | false |
openclaw/openclaw | 4,116,287,553 | I_kwDOQb6kR871WZBB | 52,312 | https://github.com/openclaw/openclaw/issues/52312 | https://api.github.com/repos/openclaw/openclaw/issues/52312 | [Config] gateway.controlUi.allowedOrigins is not a valid schema key (v0.1.8-fix.3) | ## ??????
`gateway.controlUi.allowedOrigins` ??schema ????????'unrecognized key'?????????????????n
## ???
- OpenClaw ?????.1.8-fix.3
- Node.js ?????24.14.0
- ????????indows
## ??????
1. ??? `gateway.controlUi.allowedOrigins: ["null"]` ??`{}`
2. ??? `openclaw config validate` ?????
3. ??? `openclaw-cn doctor --fix` ???... | closed | completed | true | 3 | [] | [] | 2026-03-22T15:08:46Z | 2026-04-30T04:48:49Z | 2026-04-25T05:56:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | arxir | 50,477,090 | MDQ6VXNlcjUwNDc3MDkw | User | false |
openclaw/openclaw | 4,116,320,554 | I_kwDOQb6kR871WhEq | 52,314 | https://github.com/openclaw/openclaw/issues/52314 | https://api.github.com/repos/openclaw/openclaw/issues/52314 | Feature: pre-compaction memory flush hook for agents | ## Summary
Expose a pre-compaction callback so agents can flush durable notes to disk before context is summarized/compacted.
## Problem
When a long-running session approaches context window limits, OpenClaw silently compacts history. Any state the agent was tracking in-session that wasn't already written to disk is... | closed | completed | true | 1 | [] | [] | 2026-03-22T15:22:13Z | 2026-04-30T04:48:48Z | 2026-04-25T05:56:23Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | robertvanstedum | 65,718,279 | MDQ6VXNlcjY1NzE4Mjc5 | User | false |
openclaw/openclaw | 4,116,323,739 | I_kwDOQb6kR871Wh2b | 52,317 | https://github.com/openclaw/openclaw/issues/52317 | https://api.github.com/repos/openclaw/openclaw/issues/52317 | System prompt completely missing after /new — workspace files not injected, silent failure | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
After editing 4 workspace files (AGENTS.md, HEARTBEAT.md, TOOLS.md, SOUL.md) via tool calls in a session and running `/new`, the new session started with a **completely empty system prompt** — no workspace files, no tool definitions, no per... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T15:23:48Z | 2026-04-30T04:48:47Z | 2026-04-25T05:56:29Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | qiqi-victor | 242,737,599 | U_kgDODnfhvw | User | false |
openclaw/openclaw | 4,116,335,885 | I_kwDOQb6kR871Wk0N | 52,323 | https://github.com/openclaw/openclaw/issues/52323 | https://api.github.com/repos/openclaw/openclaw/issues/52323 | Cron isolated sessions should display job name as session label in Control UI | ## Problem
When cron jobs create isolated sessions (agentTurn), the Control UI dropdown shows the session key (e.g., `cron:<uuid>`) instead of a human-readable name. Users cannot distinguish which cron job a session belongs to without cross-referencing the UUID.
## Current Behavior
The Control UI session selector sh... | closed | completed | true | 3 | [] | [] | 2026-03-22T15:30:00Z | 2026-04-30T04:48:47Z | 2026-04-25T05:56:36Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Gyrych | 42,962,689 | MDQ6VXNlcjQyOTYyNjg5 | User | false |
openclaw/openclaw | 4,116,367,675 | I_kwDOQb6kR871Wsk7 | 52,333 | https://github.com/openclaw/openclaw/issues/52333 | https://api.github.com/repos/openclaw/openclaw/issues/52333 | openclaw status: Gateway probe reports unreachable on loopback (missing scope operator.read) | ## Description
`openclaw status` always shows `unreachable (missing scope: operator.read)` on the Gateway line, even though the gateway is fully functional. All other commands work correctly:
- `openclaw health` — OK (channels connected)
- `openclaw gateway status` — `RPC probe: ok`
- `openclaw devices list` — paired... | closed | completed | true | 1 | [] | [] | 2026-03-22T15:44:58Z | 2026-04-30T04:48:44Z | 2026-04-25T05:56:55Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | chancheuklap | 252,303,303 | U_kgDODwnXxw | User | false |
openclaw/openclaw | 4,116,375,051 | I_kwDOQb6kR871WuYL | 52,338 | https://github.com/openclaw/openclaw/issues/52338 | https://api.github.com/repos/openclaw/openclaw/issues/52338 | [CLI] Local operator RPCs still lose operator.read after 2026.3.12 shared-auth hardening | ## Summary
On a Linux loopback-only gateway running OpenClaw `2026.3.13`, several local CLI/operator RPC paths still fail with `missing scope: operator.read` or WebSocket close `1000 normal closure`, even though the local paired device token was explicitly repaired to include both `operator.read` and `operator.admin`.... | closed | completed | true | 3 | [] | [] | 2026-03-22T15:48:20Z | 2026-04-30T04:48:43Z | 2026-04-25T05:57:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lightrao | 20,747,578 | MDQ6VXNlcjIwNzQ3NTc4 | User | false |
openclaw/openclaw | 4,327,257,795 | I_kwDOQb6kR88AAAABAey2ww | 71,420 | https://github.com/openclaw/openclaw/issues/71420 | https://api.github.com/repos/openclaw/openclaw/issues/71420 | [Bug]: Bundled runtime deps repeatedly reinstall / stall when packaged `/app` is non-writable | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
OpenClaw expects writable /app in containers, which is often not the case in production-grade Kubernetes environments. Observed under OpenShift restricted SCC, and likely relevant to Kubernetes/container deploy... | closed | completed | true | 3 | [
"bug"
] | [] | 2026-04-25T05:57:06Z | 2026-04-28T05:24:21Z | 2026-04-25T07:11:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Conan-Scott | 271,909,525 | U_kgDOEDUClQ | User | false |
openclaw/openclaw | 4,116,380,043 | I_kwDOQb6kR871WvmL | 52,339 | https://github.com/openclaw/openclaw/issues/52339 | https://api.github.com/repos/openclaw/openclaw/issues/52339 | [Bug]: OPENCLAW_CONFIG_DIR defaults to /root/.openclaw causing silent write failures for non-root users | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The default OPENCLAW_CONFIG_DIR=/root/.openclaw in .env causes the container's node user (uid 1000) to fail silently when writing config, session, and channel data. On a standard Linux setup where Docker is run
by a non-root user, /root/... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T15:50:50Z | 2026-04-30T04:48:43Z | 2026-04-25T05:57:09Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | bfg-jthornton | 250,869,917 | U_kgDODvP4nQ | User | false |
openclaw/openclaw | 4,116,387,516 | I_kwDOQb6kR871Wxa8 | 52,344 | https://github.com/openclaw/openclaw/issues/52344 | https://api.github.com/repos/openclaw/openclaw/issues/52344 | [Bug]: TELEGRAM_BOT_TOKEN env override is ignored on doctor/update when config uses exec SecretRef | ## Summary
`openclaw doctor` and `openclaw update` still fail on macOS when `channels.telegram.botToken` is an `exec` SecretRef, even if `TELEGRAM_BOT_TOKEN` is already exported in the shell before the command runs.
This appears to violate the documented "env wins" behavior for Telegram credentials and makes it impos... | closed | completed | true | 1 | [] | [] | 2026-03-22T15:54:29Z | 2026-04-30T04:48:42Z | 2026-04-25T05:57:15Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ge0-AuroraBot | 258,684,804 | U_kgDOD2s3hA | User | false |
openclaw/openclaw | 4,116,387,858 | I_kwDOQb6kR871WxgS | 52,345 | https://github.com/openclaw/openclaw/issues/52345 | https://api.github.com/repos/openclaw/openclaw/issues/52345 | [Feature]: clarify persistent lane vs thread-bound session UX | ### Summary
Clarify the sessions_spawn / persistent-session UX so users can reliably understand when they are creating a persistent lane, a thread-bound lane, or a named session.
### Problem to solve
The current wording around persistent sessions, thread-bound sessions, lanes, and named sessions makes the mental mod... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-22T15:54:40Z | 2026-04-30T04:48:41Z | 2026-04-25T05:57:22Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | edubur | 212,682,418 | U_kgDODK1Gsg | User | false |
openclaw/openclaw | 4,116,392,154 | I_kwDOQb6kR871Wyja | 52,347 | https://github.com/openclaw/openclaw/issues/52347 | https://api.github.com/repos/openclaw/openclaw/issues/52347 | [Bug]: memory_search fails - Gemini embedding model unsupported | memory_search returns 404: models/embedding-001 is not supported by Gemini API. Hardcoded model name is outdated. Update to text-embedding-004 or equivalent. | closed | not_planned | true | 3 | [] | [] | 2026-03-22T15:56:53Z | 2026-04-30T04:48:40Z | 2026-04-25T05:57:28Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dapzthelegend | 47,679,048 | MDQ6VXNlcjQ3Njc5MDQ4 | User | false |
openclaw/openclaw | 4,214,026,366 | I_kwDOQb6kR877LPB- | 62,121 | https://github.com/openclaw/openclaw/issues/62121 | https://api.github.com/repos/openclaw/openclaw/issues/62121 | DeepSeek preamble text leaks to Telegram after 3.13 → 4.5 upgrade (untagged assistant text bypasses commentary filter) | ## Description
After upgrading from 2026.3.13 to 2026.4.5, DeepSeek (`deepseek-chat`) cron jobs that deliver to Telegram are leaking preamble/reasoning text ("Let me fetch the data...", "Based on my research...", "Now I have enough information...") that was not visible in 3.13.
## Environment
- OpenClaw: 2026.4.5 (3... | open | null | false | 3 | [] | [] | 2026-04-06T20:56:02Z | 2026-05-02T07:02:08Z | null | NONE | null | 20260502T090030Z | 2026-05-02T09:00:30Z | tlnini-afk | 270,210,840 | U_kgDOEBsXGA | User | false |
openclaw/openclaw | 4,116,447,087 | I_kwDOQb6kR871W_9v | 52,362 | https://github.com/openclaw/openclaw/issues/52362 | https://api.github.com/repos/openclaw/openclaw/issues/52362 | claude-cli backend: assistant responses not persisted to session .jsonl, webchat history lost on reload | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
When using claude-cli as the model backend in OpenClaw, assistant responses are delivered in real time but never persisted to the session .jsonl file, causing the webchat history to be lost on page reload.
### Steps to reproduce
Configure... | closed | completed | true | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T16:22:26Z | 2026-04-30T04:48:40Z | 2026-04-25T05:57:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | MRosende | 93,163,460 | U_kgDOBY2PxA | User | false |
openclaw/openclaw | 4,116,488,763 | I_kwDOQb6kR871XKI7 | 52,368 | https://github.com/openclaw/openclaw/issues/52368 | https://api.github.com/repos/openclaw/openclaw/issues/52368 | exec tool loses requester session context due to sessionKey / agentSessionKey mismatch | ## Summary
While debugging a formal-task origin-pinning regression in Telegram topic sessions, I found a lower-level runtime bug in the `exec` tool creation path.
`createOpenClawTools(...)` consistently passes `agentSessionKey` to most tools, but the `exec` tool was constructed with `sessionKey: options?.sessionKey` i... | closed | completed | true | 2 | [] | [] | 2026-03-22T16:42:17Z | 2026-04-30T04:48:39Z | 2026-04-25T05:57:41Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | clawbotbkk | 270,162,996 | U_kgDOEBpcNA | User | false |
openclaw/openclaw | 4,116,491,997 | I_kwDOQb6kR871XK7d | 52,370 | https://github.com/openclaw/openclaw/issues/52370 | https://api.github.com/repos/openclaw/openclaw/issues/52370 | HEARTBEAT_OK response leaks to Slack DM instead of being stripped | ## Bug
When a heartbeat fires concurrently with an active Slack DM conversation, the `HEARTBEAT_OK` response is delivered to the user's DM instead of being silently consumed.
## Expected Behavior
`HEARTBEAT_OK` responses should always be stripped before reaching the user, per the [heartbeat docs](https://docs.opencl... | closed | completed | true | 3 | [] | [] | 2026-03-22T16:43:39Z | 2026-04-30T04:48:37Z | 2026-04-25T05:57:54Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | scottyallen | 73,208 | MDQ6VXNlcjczMjA4 | User | false |
openclaw/openclaw | 4,116,520,040 | I_kwDOQb6kR871XRxo | 52,383 | https://github.com/openclaw/openclaw/issues/52383 | https://api.github.com/repos/openclaw/openclaw/issues/52383 | feat: configurable fallbackAfterRetries for model failover under sustained 429s | ## Summary
Model fallback is configured correctly (e.g., Gemini → Claude → GPT-4o) but the retry count before switching is too high. Under sustained 429 flooding, the platform burns multiple retry cycles against the failing provider before switching, creating an unresponsive window that can last 30-60+ seconds.
## Ob... | closed | completed | true | 2 | [] | [] | 2026-03-22T16:56:15Z | 2026-04-30T04:48:36Z | 2026-04-25T05:58:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | barry-p5cc | 258,459,530 | U_kgDOD2fHig | User | false |
openclaw/openclaw | 4,116,555,052 | I_kwDOQb6kR871XaUs | 52,390 | https://github.com/openclaw/openclaw/issues/52390 | https://api.github.com/repos/openclaw/openclaw/issues/52390 | [Bug]: message:sent internal hook not firing for Telegram group deliveries (missing sessionKey) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Description
The message:sent internal hook event is never emitted when a bot delivers a message to a Telegram group (including forum topic chats). The message:received and message:preprocessed events fire correctly for the same conversatio... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-22T17:09:43Z | 2026-04-30T04:48:35Z | 2026-04-25T05:58:14Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ISE76R | 190,644,940 | U_kgDOC10CzA | User | false |
openclaw/openclaw | 4,116,593,874 | I_kwDOQb6kR871XjzS | 52,398 | https://github.com/openclaw/openclaw/issues/52398 | https://api.github.com/repos/openclaw/openclaw/issues/52398 | [Bug]: totalTokens null crash during API instability | ### Bug type
Regression (worked before, now fails)
### Summary
The main session crashes when the Anthropic API returns a malformed response with a null `totalTokens` value during API instability / degraded service windows. Expected: graceful null-check + retry or skip. References are in minified dist files so userla... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-22T17:27:36Z | 2026-04-30T04:48:34Z | 2026-04-25T05:58:21Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wow-leeroy-jenkins05 | 266,177,925 | U_kgDOD92NhQ | User | false |
openclaw/openclaw | 4,116,688,426 | I_kwDOQb6kR871X64q | 52,411 | https://github.com/openclaw/openclaw/issues/52411 | https://api.github.com/repos/openclaw/openclaw/issues/52411 | [Feature]: Expose runId in agent hook context (before_agent_start, before_prompt_build, agent_end) | ### Summary
Expose the current agent run's `runId` in the `PluginHookAgentContext` passed to agent lifecycle hooks (`before_agent_start`, `before_prompt_build`, `agent_end`, etc.).
### Problem to solve
Plugins that need to track or manage individual agent runs (e.g., cancellation, deduplication, observability) curre... | closed | completed | true | 1 | [] | [] | 2026-03-22T18:09:30Z | 2026-04-30T04:48:33Z | 2026-04-25T05:58:27Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | livingghost | 14,040,213 | MDQ6VXNlcjE0MDQwMjEz | User | false |
openclaw/openclaw | 4,116,756,022 | I_kwDOQb6kR871YLY2 | 52,424 | https://github.com/openclaw/openclaw/issues/52424 | https://api.github.com/repos/openclaw/openclaw/issues/52424 | cron create/list fails with 'gateway closed (1000)' handshake timeout on Windows | ## 🐛 Bug Description
`openclaw cron` commands (both create and list) fail with `gateway closed (1000: normal closure)` immediately after gateway restart, even when gateway is confirmed running and responsive.
## Environment
- OS: Windows (NT 10.0.26200)
- OpenClaw: 2026.3.13 (61d171a)
- Node.js: v24.13.0
- Installa... | closed | completed | true | 5 | [] | [] | 2026-03-22T18:37:04Z | 2026-04-30T04:48:33Z | 2026-04-25T05:58:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | cuizhou6463-lang | 263,814,057 | U_kgDOD7l7qQ | User | false |
openclaw/openclaw | 4,116,822,963 | I_kwDOQb6kR871Ybuz | 52,438 | https://github.com/openclaw/openclaw/issues/52438 | https://api.github.com/repos/openclaw/openclaw/issues/52438 | bug: CI broken on main — TypeScript error in zod-schema.core.ts (thinkingFormat: 'openrouter') | ## Description
All CI checks are currently failing on `main` for every open PR. The root cause is a TypeScript type error:
```
src/config/zod-schema.core.ts(221,3): error TS2344: Type 'ModelCompatConfig | undefined' does not satisfy the constraint...
Types of property 'thinkingFormat' are incompatible.
Type 'Su... | closed | completed | true | 2 | [] | [] | 2026-03-22T18:59:58Z | 2026-04-30T04:48:31Z | 2026-04-25T05:58:48Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | BrainMode | 128,379,663 | U_kgDOB6brDw | User | false |
openclaw/openclaw | 4,116,828,098 | I_kwDOQb6kR871Yc_C | 52,439 | https://github.com/openclaw/openclaw/issues/52439 | https://api.github.com/repos/openclaw/openclaw/issues/52439 | exec.approval.resolve fails with unknown or expired approval id in Control UI / gateway flow | # exec.approval.resolve fails with `unknown or expired approval id` in Control UI / gateway flow
## Version / environment
- OpenClaw: `2026.3.13 (61d171a)`
- Surface: `openclaw-control-ui`
- Transport: Control UI → Gateway WebSocket
- Gateway service running locally
## Symptom
`exec.approval.waitDecision` succeeds, a... | closed | completed | true | 5 | [] | [] | 2026-03-22T19:01:47Z | 2026-04-30T04:48:30Z | 2026-04-25T05:58:54Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Marsero | 18,037,836 | MDQ6VXNlcjE4MDM3ODM2 | User | false |
openclaw/openclaw | 4,116,835,536 | I_kwDOQb6kR871YezQ | 52,440 | https://github.com/openclaw/openclaw/issues/52440 | https://api.github.com/repos/openclaw/openclaw/issues/52440 | ACP sessions killed by gateway restart (SIGUSR1) | ## Problem
When the gateway restarts (via `config.patch`, adding agents, or any SIGUSR1-triggered reload), all running ACP sessions are terminated because the ACP runtime processes are children of the gateway process.
This means **any long-running ACP task** (e.g., a Codex code review of 845 files taking 15-30+ minut... | closed | completed | true | 4 | [] | [] | 2026-03-22T19:04:22Z | 2026-04-30T04:48:29Z | 2026-04-25T05:59:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | skillz-xx | 837,936 | MDQ6VXNlcjgzNzkzNg== | User | false |
openclaw/openclaw | 4,116,908,934 | I_kwDOQb6kR871YwuG | 52,452 | https://github.com/openclaw/openclaw/issues/52452 | https://api.github.com/repos/openclaw/openclaw/issues/52452 | ACP sessions silently fail/die without completion events | ## Problem
ACP sessions spawned via `sessions_spawn(runtime='acp')` frequently die without emitting any completion or error event back to the parent session. The parent agent has no way to know the task failed — it just never hears back.
In my testing (OpenAI Codex CLI via acpx backend), 3 out of 3 ACP runs failed:
... | closed | completed | true | 9 | [] | [
"vincentkoc"
] | 2026-03-22T19:34:13Z | 2026-04-30T04:48:29Z | 2026-04-25T05:59:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | skillz-xx | 837,936 | MDQ6VXNlcjgzNzkzNg== | User | false |
openclaw/openclaw | 4,116,915,367 | I_kwDOQb6kR871YySn | 52,453 | https://github.com/openclaw/openclaw/issues/52453 | https://api.github.com/repos/openclaw/openclaw/issues/52453 | Gateway WebSocket rejects all inbound client connections with handshake timeout (2026.3.13) | # GitHub Issue Draft: Gateway WebSocket Handshake Timeout
---
## Title
**Gateway WebSocket rejects all inbound client connections with handshake timeout (2026.3.13)**
---
## Description
The OpenClaw gateway WebSocket server (`ws://127.0.0.1:18789`) is rejecting all inbound client connections with a handshake timeo... | closed | completed | true | 3 | [] | [] | 2026-03-22T19:37:01Z | 2026-04-30T04:48:28Z | 2026-04-25T05:59:14Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | coreydschultz | 260,803,254 | U_kgDOD4uKtg | User | false |
openclaw/openclaw | 4,116,996,554 | I_kwDOQb6kR871ZGHK | 52,464 | https://github.com/openclaw/openclaw/issues/52464 | https://api.github.com/repos/openclaw/openclaw/issues/52464 | WhatsApp messages arrive as [object Object] — root cause in sanitizeChatSendMessageInput() | ## Bug Description
WhatsApp messages intermittently arrive to the LLM as `[object Object]` instead of the actual text. The bot sees garbled input and cannot process the user's request.
Related to #20566 (locked).
## Root Cause (traced in source)
**`sanitizeChatSendMessageInput()` in `gateway-cli-CuZs0RlJ.js:10377`*... | closed | not_planned | true | 4 | [] | [] | 2026-03-22T20:12:27Z | 2026-04-30T04:48:26Z | 2026-04-25T05:59:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | frdaniel76 | 173,002,915 | U_kgDOCk_Qow | User | false |
openclaw/openclaw | 4,117,015,729 | I_kwDOQb6kR871ZKyx | 52,468 | https://github.com/openclaw/openclaw/issues/52468 | https://api.github.com/repos/openclaw/openclaw/issues/52468 | Best practice for external middleware to inject messages into a channel session? | ## Use Case
I'm building a **channel monitoring middleware** that watches a real estate agent's personal Line/WhatsApp/Email for buyer messages, then relays them to an OpenClaw agent for processing (lead tracking, draft replies, agent approval via Telegram).
The middleware monitors channels that OpenClaw can't native... | closed | completed | true | 2 | [] | [] | 2026-03-22T20:21:28Z | 2026-04-30T04:48:26Z | 2026-04-25T05:59:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mcc-weng | 228,220,534 | U_kgDODZpedg | User | false |
openclaw/openclaw | 4,117,044,059 | I_kwDOQb6kR871ZRtb | 52,471 | https://github.com/openclaw/openclaw/issues/52471 | https://api.github.com/repos/openclaw/openclaw/issues/52471 | Control UI: Model selector uses wrong provider prefix (ollama/ instead of ollama-remote/) | ## Bug Description
The Control UI model dropdown selector generates incorrect model identifiers when selecting models from non-default providers.
## Steps to Reproduce
1. Configure multiple Ollama providers in openclaw.json:
- `ollama` (local: 127.0.0.1:11434)
- `ollama-remote` (remote: 192.168.1.175:11434)
2... | closed | completed | true | 4 | [] | [] | 2026-03-22T20:34:54Z | 2026-04-30T04:48:25Z | 2026-04-25T05:59:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | wangcc61 | 81,038,721 | MDQ6VXNlcjgxMDM4NzIx | User | false |
openclaw/openclaw | 4,117,072,382 | I_kwDOQb6kR871ZYn- | 52,474 | https://github.com/openclaw/openclaw/issues/52474 | https://api.github.com/repos/openclaw/openclaw/issues/52474 | Desktop UI model dropdown prepends 'anthropic/' to all non-Anthropic models | ## Bug Report
**OpenClaw version:** 2026.3.13
**Platform:** macOS (Apple M4 Mac mini)
### Description
The model dropdown in the desktop Control UI incorrectly prepends `anthropic/` to all model names regardless of provider. This causes `sessions.patch` to fail with:
```
GatewayRequestError: model not allowed: anthr... | closed | completed | true | 5 | [] | [] | 2026-03-22T20:46:11Z | 2026-04-30T04:48:24Z | 2026-04-25T05:59:47Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | lucpritchett-hash | 270,196,743 | U_kgDOEBrgBw | User | false |
openclaw/openclaw | 4,117,205,257 | I_kwDOQb6kR871Z5EJ | 52,481 | https://github.com/openclaw/openclaw/issues/52481 | https://api.github.com/repos/openclaw/openclaw/issues/52481 | Feature Request: Release Roadmap / Changelog Preview | ## Description
It would be helpful to have a way to see upcoming releases and what features they will include, similar to a roadmap or release preview.
## Problem
Currently, users only find out about new features after a release is published. There's no way to anticipate what will be included in the next release.
#... | closed | completed | true | 1 | [] | [] | 2026-03-22T21:55:29Z | 2026-04-30T04:48:23Z | 2026-04-25T05:59:53Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | iamliba2k | 259,181,745 | U_kgDOD3LMsQ | User | false |
openclaw/openclaw | 4,117,269,702 | I_kwDOQb6kR871aIzG | 52,494 | https://github.com/openclaw/openclaw/issues/52494 | https://api.github.com/repos/openclaw/openclaw/issues/52494 | [Bug]: ui:build fails on fresh clone due to --prod removing vite | ### Bug type
Regression (worked before, now fails)
### Summary
Running `pnpm ui:build` on a fresh clone fails because the UI build script installs dependencies with `--prod`, which excludes `vite`, even though `vite` is required to run the build.
### Steps to reproduce
1. Clone the repository
2. Run `pnpm install`... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-22T22:27:14Z | 2026-04-30T04:48:23Z | 2026-04-25T05:59:59Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | K0mp0t1kk | 108,582,460 | U_kgDOBnjWPA | User | false |
openclaw/openclaw | 4,117,404,812 | I_kwDOQb6kR871apyM | 52,525 | https://github.com/openclaw/openclaw/issues/52525 | https://api.github.com/repos/openclaw/openclaw/issues/52525 | [Windows] Native file dialog invisible to browser-automation — exec+PowerShell workaround | **Environment:**
- OpenClaw 2026.3.13 (61d171a)
- Windows 10 + WSL2
- Chrome CDP on 127.0.0.1:9222
**Problem:**
When browser-automation clicks a button that opens a native
Windows file picker dialog, the dialog is invisible to the
browser tool because it renders outside the browser context.
The agent gets stuck wit... | closed | completed | true | 2 | [] | [] | 2026-03-22T23:45:56Z | 2026-04-30T04:48:22Z | 2026-04-25T06:00:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Owlock | 135,854,806 | U_kgDOCBj61g | User | false |
openclaw/openclaw | 4,117,485,575 | I_kwDOQb6kR871a9gH | 52,535 | https://github.com/openclaw/openclaw/issues/52535 | https://api.github.com/repos/openclaw/openclaw/issues/52535 | [Feature]: Image Face Recognition + identity Memory Skill (sam-faces) | ### Summary
Give ability to AI assistants to see faces in photos, automatically identify them (% confidence), and enroll new faces as unknown ones appear.
### Problem to solve
When you send an OpenClaw agent a photo, it describes what's visible but has no identity memory. It might say "a woman with red hair standi... | closed | not_planned | true | 5 | [
"enhancement"
] | [] | 2026-03-23T00:24:45Z | 2026-04-28T04:32:39Z | 2026-04-25T06:00:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jasonacox | 836,718 | MDQ6VXNlcjgzNjcxOA== | User | false |
openclaw/openclaw | 4,117,496,826 | I_kwDOQb6kR871bAP6 | 52,537 | https://github.com/openclaw/openclaw/issues/52537 | https://api.github.com/repos/openclaw/openclaw/issues/52537 | Independent agent does not inherit thinkingDefault; WeChat plugin leaks reasoning to chat | ## Problem
Two related issues when using independent agents with the WeChat (openclaw-weixin) channel:
### 1. Independent agent does not inherit `agents.defaults.thinkingDefault`
When `thinkingDefault: "off"` is set under `agents.defaults`, independent agents (created via `openclaw agents add`) do **not** inherit th... | closed | completed | true | 6 | [] | [] | 2026-03-23T00:30:06Z | 2026-04-30T04:48:20Z | 2026-04-25T06:00:34Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | nevinxie | 3,717,486 | MDQ6VXNlcjM3MTc0ODY= | User | false |
openclaw/openclaw | 4,117,506,548 | I_kwDOQb6kR871bCn0 | 52,538 | https://github.com/openclaw/openclaw/issues/52538 | https://api.github.com/repos/openclaw/openclaw/issues/52538 | [Feature]: | ### Summary
I need a refund for two CLAWBRO charges:
- 31,245 KRW
- 609,067 KRW
I cannot use the service. Please refund immediately.
### Problem to solve
I need a refund for two CLAWBRO charges:
- 31,245 KRW
- 609,067 KRW
I cannot use the service. Please refund immediately.
### Proposed solution
I need a refund fo... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-23T00:34:25Z | 2026-04-30T04:48:19Z | 2026-04-25T06:00:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | drunkenjp-dev | 270,222,828 | U_kgDOEBtF7A | User | false |
openclaw/openclaw | 4,117,567,741 | I_kwDOQb6kR871bRj9 | 52,549 | https://github.com/openclaw/openclaw/issues/52549 | https://api.github.com/repos/openclaw/openclaw/issues/52549 | The Silent Failures Problem — What the Community Is Running Into | **It worked. Or at least, that's what the logs say.**
The hardest class of failure to debug: success responses that aren't. The system confirms completion. Nothing happened.
**What the community is running into:**
- fix(core): preserve totalTokens on zero usage reports (Consolidated v2) (h/t @ryansorr in #51673)
- O... | closed | not_planned | true | 2 | [] | [] | 2026-03-23T01:00:03Z | 2026-04-30T04:48:18Z | 2026-04-25T06:00:47Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ocdlmv1 | 266,897,137 | U_kgDOD-iG8Q | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.