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 | 4,170,240,727 | I_kwDOQb6kR874kNLX | 57,683 | https://github.com/openclaw/openclaw/issues/57683 | https://api.github.com/repos/openclaw/openclaw/issues/57683 | [Bug]: 400 gemini-2.5-pro thinking_budget to 0 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
LLM error: { "error": { "code": 400, "message": "Unable to submit request because The model does not support setting thinking_budget to 0.. Learn more: https://cloud.google.com/vertex-ai/generative-ai/docs/mode... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T12:13:17Z | 2026-04-28T04:33:00Z | 2026-04-25T12:17:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | JieRenChen | 18,524,574 | MDQ6VXNlcjE4NTI0NTc0 | User | false |
openclaw/openclaw | 4,170,253,793 | I_kwDOQb6kR874kQXh | 57,685 | https://github.com/openclaw/openclaw/issues/57685 | https://api.github.com/repos/openclaw/openclaw/issues/57685 | bug: edit multi-edit validation rejects edits[]; control-ui can stall on stream failures | ## Summary
Two reliability issues seen in production-like use of OpenClaw Control UI:
1) `edit` tool multi-edit mode (`edits[]`) appears broken by parameter validation.
2) Control UI sessions can appear to "stall / no reply" during model stream failures and reconnect windows.
## Environment
- OpenClaw: 2026.3.28 (f9b... | closed | completed | true | 2 | [] | [] | 2026-03-30T12:15:27Z | 2026-04-28T05:17:53Z | 2026-04-25T12:17:09Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ThirteenStrokes | 31,041,297 | MDQ6VXNlcjMxMDQxMjk3 | User | false |
openclaw/openclaw | 4,170,341,987 | I_kwDOQb6kR874kl5j | 57,688 | https://github.com/openclaw/openclaw/issues/57688 | https://api.github.com/repos/openclaw/openclaw/issues/57688 | [Bug]: cron add fails with "pairing required" - no way to approve CLI pairing in OpenClaw 2026.3.28 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
**Body**:
```
## Bug Description
`openclaw cron add` always fails with error:
```
gateway closed (1008): pairing required
```
This happens regardless of how the command is invoked — CLI terminal, web UI, or via the agent to... | closed | completed | true | 3 | [
"bug"
] | [] | 2026-03-30T12:29:15Z | 2026-04-28T05:17:52Z | 2026-04-25T12:17:16Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chenting127 | 56,878,480 | MDQ6VXNlcjU2ODc4NDgw | User | false |
openclaw/openclaw | 4,170,579,753 | I_kwDOQb6kR874lf8p | 57,701 | https://github.com/openclaw/openclaw/issues/57701 | https://api.github.com/repos/openclaw/openclaw/issues/57701 | Control UI loses operator scopes when dangerouslyDisableDeviceAuth is enabled | ## Summary
When `gateway.controlUi.dangerouslyDisableDeviceAuth=true` is enabled, the Control UI websocket handshake can succeed but the resulting connection loses all operator scopes.
In practice this makes the UI connect successfully while every operator call fails with errors like:
- `GatewayRequestError: missing... | closed | completed | true | 2 | [] | [] | 2026-03-30T13:03:27Z | 2026-04-28T05:17:52Z | 2026-04-25T12:17:23Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | AboutLabs | 8,488,899 | MDQ6VXNlcjg0ODg4OTk= | User | false |
openclaw/openclaw | 4,170,585,214 | I_kwDOQb6kR874lhR- | 57,704 | https://github.com/openclaw/openclaw/issues/57704 | https://api.github.com/repos/openclaw/openclaw/issues/57704 | [Resource] 11 production-ready agent packs with 92 skills (open source, free) | Hi! I maintain [awesome-openclaw-agent-packs](https://github.com/clawpod-app/awesome-openclaw-agent-packs), an open-source collection of **11 pre-configured agent packs** covering sales, engineering, data analytics, marketing, legal, finance, product management, HR, customer support, design, and operations.
Each pack ... | closed | not_planned | true | 1 | [] | [] | 2026-03-30T13:04:13Z | 2026-04-28T05:17:51Z | 2026-04-25T12:17:30Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | clawpod-app | 271,940,103 | U_kgDOEDV6Bw | User | false |
openclaw/openclaw | 4,170,637,366 | I_kwDOQb6kR874luA2 | 57,706 | https://github.com/openclaw/openclaw/issues/57706 | https://api.github.com/repos/openclaw/openclaw/issues/57706 | Proposal: rename sessions.compact → sessions.compactTrim to prevent data loss confusion | ## Problem
The gateway RPC method `sessions.compact` and the user-facing `/compact` command do fundamentally different things, but share the same name:
| | `sessions.compact` (RPC) | `/compact` (slash command) |
|---|---|---|
| **Mechanism** | Raw line truncation: `lines.slice(-maxLines)` | AI-powered semantic compac... | closed | completed | true | 1 | [] | [] | 2026-03-30T13:11:42Z | 2026-04-28T05:17:50Z | 2026-04-25T12:17:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Kaspre | 36,520,309 | MDQ6VXNlcjM2NTIwMzA5 | User | false |
openclaw/openclaw | 4,170,832,749 | I_kwDOQb6kR874mdtt | 57,718 | https://github.com/openclaw/openclaw/issues/57718 | https://api.github.com/repos/openclaw/openclaw/issues/57718 | WhatsApp creds.json repeatedly restored from backup on startup, even after fresh re-pair | On the current OpenClaw release, the WhatsApp channel links successfully and remains operational, but on restart it repeatedly reports that `creds.json` was corrupted and restored from backup.
This behavior persists even after a fresh WhatsApp re-pair.
Observed behavior:
- WhatsApp starts successfully
- startup repor... | closed | completed | true | 1 | [] | [] | 2026-03-30T13:40:01Z | 2026-04-28T05:17:49Z | 2026-04-25T12:17:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jtyun | 16,197,533 | MDQ6VXNlcjE2MTk3NTMz | User | false |
openclaw/openclaw | 4,170,873,906 | I_kwDOQb6kR874mnwy | 57,724 | https://github.com/openclaw/openclaw/issues/57724 | https://api.github.com/repos/openclaw/openclaw/issues/57724 | Discord native slash commands silently fail - InteractionEventListener processes INTERACTION_CREATE but never responds | ## Describe the bug
Discord native slash commands (e.g. `/status`, `/help`) receive `INTERACTION_CREATE` events but never produce a response. The gateway logs show the interaction listener processing the event (1-1.5s) but no session is created, no model call is made, no error is logged, and no response is sent back t... | closed | completed | true | 1 | [] | [] | 2026-03-30T13:45:51Z | 2026-04-28T05:17:49Z | 2026-04-25T12:17:51Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | zeejaytan | 217,020,377 | U_kgDODO932Q | User | false |
openclaw/openclaw | 4,171,027,400 | I_kwDOQb6kR874nNPI | 57,731 | https://github.com/openclaw/openclaw/issues/57731 | https://api.github.com/repos/openclaw/openclaw/issues/57731 | [Bug]: Discord plugin aggressively drops connections (stale-socket) causing delayed and duplicated messages | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Since the latest update, the Discord plugin aggressively drops its websocket connection, resulting in constant `stale-socket` restarts from the `health-monitor`. When the socket goes stale, the bot stops receiving Discord e... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-30T14:05:21Z | 2026-04-28T05:17:48Z | 2026-04-25T12:18:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | agentandyai | 7,111,080 | MDQ6VXNlcjcxMTEwODA= | User | false |
openclaw/openclaw | 4,171,255,258 | I_kwDOQb6kR874oE3a | 57,742 | https://github.com/openclaw/openclaw/issues/57742 | https://api.github.com/repos/openclaw/openclaw/issues/57742 | [Bug]: ACP Codex replies are routed to main session, while thread binding is unavailable in current TUI session | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Hi team,
I’m reporting what looks like a bug / broken workflow when using OpenClaw + ACP/ACPX + Codex from the current OpenClaw TUI session.
## Summary
I want to use Codex interactively from the current TUI session, with... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-30T14:30:48Z | 2026-04-28T05:17:47Z | 2026-04-25T12:18:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rainzhao | 13,361,732 | MDQ6VXNlcjEzMzYxNzMy | User | false |
openclaw/openclaw | 4,171,295,441 | I_kwDOQb6kR874oOrR | 57,743 | https://github.com/openclaw/openclaw/issues/57743 | https://api.github.com/repos/openclaw/openclaw/issues/57743 | Telegram: getUpdates polling stalls for 90-110s intermittently despite healthy network and IPv4 forced | ## Environment
- OpenClaw version: `2026.3.28`
- OS: Debian 13, Linux 6.12.74
- Node: v22.22.0
- IPv4 forced via systemd drop-in: `NODE_OPTIONS=--dns-result-order=ipv4first`
## Problem
The Telegram `getUpdates` long-poll socket stalls for **90–110 seconds** intermittently, even though:
- Direct `curl -4 https://api.t... | closed | completed | true | 1 | [] | [] | 2026-03-30T14:35:46Z | 2026-04-28T05:17:46Z | 2026-04-25T12:18:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rayrage213 | 261,946,476 | U_kgDOD5z8bA | User | false |
openclaw/openclaw | 4,171,370,213 | I_kwDOQb6kR874og7l | 57,753 | https://github.com/openclaw/openclaw/issues/57753 | https://api.github.com/repos/openclaw/openclaw/issues/57753 | [Bug]: anthropic-messages SSE state machine fails to reset between tool-use loop iterations — "Unexpected event order, got message_start before receiving message_stop" | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from 2026.3.13 to 2026.3.28, every request to kimi-coding/k2p5 via the anthropic-messages API protocol fails with "Unexpected event order, got message_start before receiving message_stop". The error occurs 10... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-30T14:45:05Z | 2026-04-28T05:17:46Z | 2026-04-25T12:18:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | flybetter2005 | 79,260,285 | MDQ6VXNlcjc5MjYwMjg1 | User | false |
openclaw/openclaw | 4,171,508,596 | I_kwDOQb6kR874pCt0 | 57,768 | https://github.com/openclaw/openclaw/issues/57768 | https://api.github.com/repos/openclaw/openclaw/issues/57768 | [Feature]: adopt some Hermes-style architectural ideas for longer-term learning and RL-driven improvement | ### Summary
Hi, I’m an OpenClaw user.
I really like OpenClaw overall, but after looking at Hermes Agent, I felt some of its architectural ideas could improve OpenClaw as well.
In particular, I think OpenClaw could benefit from:
- a clearer long-term improvement loop
- stronger integration between user feedback and a... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-30T15:01:55Z | 2026-04-28T05:17:45Z | 2026-04-25T12:18:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | junyoung6215 | 168,890,027 | U_kgDOChEOqw | User | false |
openclaw/openclaw | 4,171,546,530 | I_kwDOQb6kR874pL-i | 57,774 | https://github.com/openclaw/openclaw/issues/57774 | https://api.github.com/repos/openclaw/openclaw/issues/57774 | [Bug]: 子agent的session没有正确退出，导致死循环 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Subagent session结束后，系统重复推送相同的completion announcement事件。
同一个task_id（minimaxi-image-skill）被反复推送多次完成事件。
### Steps to reproduce
1. 通过sessions_send向subagent发任务
2. 任务完成后，subagent发送announcement
3. announcement被推送后，s... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T15:06:28Z | 2026-04-28T05:17:44Z | 2026-04-25T12:18:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bai7988 | 263,560,998 | U_kgDOD7WfJg | User | false |
openclaw/openclaw | 4,172,012,118 | I_kwDOQb6kR874q9pW | 57,812 | https://github.com/openclaw/openclaw/issues/57812 | https://api.github.com/repos/openclaw/openclaw/issues/57812 | LiveSessionModelSwitchError overrides model failover, creating infinite retry loop | ## Description
When the primary model (e.g., `anthropic/claude-opus-4-6`) hits rate limits and all auth profiles enter cooldown, the failover engine correctly decides to fall back to the next model in `agents.defaults.model.fallbacks` (e.g., `openai-codex/gpt-5.4`). However, the fallback candidate never actually execu... | closed | completed | true | 1 | [] | [] | 2026-03-30T16:08:41Z | 2026-04-28T05:17:43Z | 2026-04-25T12:19:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bill492 | 257,063,863 | U_kgDOD1J7tw | User | false |
openclaw/openclaw | 4,172,015,517 | I_kwDOQb6kR874q-ed | 57,813 | https://github.com/openclaw/openclaw/issues/57813 | https://api.github.com/repos/openclaw/openclaw/issues/57813 | [Bug]: UPDATE - Root Cause Found NEXTCLOUD TALK | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
Nextcloud Talk's BotController.php are having some differences. This was a closed issue, but apparently it's not.
### Steps to reproduce
The HMAC signature generation in send.ts is incorrect. Signs full JSON body instead o... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-03-30T16:09:05Z | 2026-04-28T05:17:43Z | 2026-04-25T12:19:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Strellnikov | 10,959,245 | MDQ6VXNlcjEwOTU5MjQ1 | User | false |
openclaw/openclaw | 4,172,043,594 | I_kwDOQb6kR874rFVK | 57,814 | https://github.com/openclaw/openclaw/issues/57814 | https://api.github.com/repos/openclaw/openclaw/issues/57814 | [Bug] [agent/embedded] [ws-stream] WebSocket connect failed 500 on openai-codex/gpt-5.4, not on other models | ## Summary
When using `openai-codex/gpt-5.4` as the model, the embedded agent's WebSocket streaming connection consistently fails with HTTP 500. The system falls back to HTTP, but the 500 errors continue to appear in `gateway.err.log` repeatedly.
When switching to other models (qianfan, glm-5, kimibd), the ws-stream ... | closed | completed | true | 5 | [] | [] | 2026-03-30T16:12:12Z | 2026-04-28T05:17:42Z | 2026-04-25T12:19:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | yijingy224 | 207,885,400 | U_kgDODGQUWA | User | false |
openclaw/openclaw | 4,172,078,793 | I_kwDOQb6kR874rN7J | 57,817 | https://github.com/openclaw/openclaw/issues/57817 | https://api.github.com/repos/openclaw/openclaw/issues/57817 | [Bug] contextWindow incorrectly merged/shared across providers when using models.mode "merge" | ## Bug Description
When using `models.mode: "merge"` with multiple providers (ollama + minimax), the `contextWindow` value appears to be incorrectly tracked or shared across providers. Even though each model has its own `contextWindow` defined in `models.providers.<provider>.models`, the system uses the wrong value.
... | closed | completed | true | 1 | [] | [] | 2026-03-30T16:16:34Z | 2026-04-28T05:17:41Z | 2026-04-25T12:19:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | shinfon | 92,320,216 | U_kgDOBYCx2A | User | false |
openclaw/openclaw | 4,172,167,327 | I_kwDOQb6kR874rjif | 57,821 | https://github.com/openclaw/openclaw/issues/57821 | https://api.github.com/repos/openclaw/openclaw/issues/57821 | [Bug]: Cannot read properties of undefined (reading 'push') with third-party Anthropic-compatible provider returning non-standard content blocks | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Every agent run fails with `Cannot read properties of undefined (reading 'push')` when using a third-party Anthropic-compatible relay provider that returns non-standard content blocks containing an extra `"id": ""` field.
#... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-30T16:25:27Z | 2026-04-28T05:17:40Z | 2026-04-25T12:19:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | PGokc | 67,835,432 | MDQ6VXNlcjY3ODM1NDMy | User | false |
openclaw/openclaw | 4,172,167,596 | I_kwDOQb6kR874rjms | 57,822 | https://github.com/openclaw/openclaw/issues/57822 | https://api.github.com/repos/openclaw/openclaw/issues/57822 | [ Bug ]: Web UI agent dropdown shows agent.id instead of identity.name | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When multiple agents are configured with distinct `identity.name` values in `openclaw.json`, the Web UI agent switcher dropdown displays `agent.id` (e.g., "main", "dev", "writer") instead of the human-readable ... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-30T16:25:29Z | 2026-04-28T05:17:39Z | 2026-04-25T12:19:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | write31bug | 54,340,965 | MDQ6VXNlcjU0MzQwOTY1 | User | false |
openclaw/openclaw | 4,172,248,337 | I_kwDOQb6kR874r3UR | 57,827 | https://github.com/openclaw/openclaw/issues/57827 | https://api.github.com/repos/openclaw/openclaw/issues/57827 | [Bug]: OpenClaw 2026.3.28 `config.openFile` / `OPENCLAW_CONFIG_PATH Command Injection | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The OpenClaw Gateway component (v2026.3.28) is vulnerable to an Unauthenticated Remote Code Execution (RCE) via the config.openFile method due to insecure usage of the Node.js child_process.exec() function.
##... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T16:35:40Z | 2026-04-28T05:17:38Z | 2026-04-25T12:19:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | woshizhuchenguang | 71,500,166 | MDQ6VXNlcjcxNTAwMTY2 | User | false |
openclaw/openclaw | 4,172,321,002 | I_kwDOQb6kR874sJDq | 57,841 | https://github.com/openclaw/openclaw/issues/57841 | https://api.github.com/repos/openclaw/openclaw/issues/57841 | Anthropic Opus breaks on 2026.3.28 with subscription/setup-token auth | ## Environment
- **OpenClaw version:** 2026.3.28 (f9b1079)
- **OS:** macOS 26.3.2 (arm64)
- **Auth method:** Anthropic setup-token (`sk-ant-oat01-...`, generated via `claude setup-token` from Claude Code CLI v2.1.87)
- **Subscription:** Max plan
## Problem
After upgrading from 2026.3.24 to 2026.3.28, Anthropic Opus (`... | closed | completed | true | 1 | [] | [] | 2026-03-30T16:45:57Z | 2026-04-28T05:17:37Z | 2026-04-25T12:20:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | arisgysel-design | 259,428,272 | U_kgDOD3aPsA | User | false |
openclaw/openclaw | 4,172,474,540 | I_kwDOQb6kR874suis | 57,854 | https://github.com/openclaw/openclaw/issues/57854 | https://api.github.com/repos/openclaw/openclaw/issues/57854 | Gateway crashes on unhandled fetch rejection (~12 min intervals, 27+ restarts/day) | ## Bug Description
The gateway process crashes repeatedly due to an unhandled promise rejection from a failed `fetch` call. The process exits and systemd restarts it, but in-flight sessions are killed without recovery (related to #57851).
## Environment
- **Clawdbot version:** 2026.1.24-3
- **Node.js:** v22.22.0
- *... | closed | completed | true | 2 | [] | [] | 2026-03-30T17:05:37Z | 2026-04-28T05:17:36Z | 2026-04-25T12:20:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nrubioru | 256,029,441 | U_kgDOD0KzAQ | User | false |
openclaw/openclaw | 4,172,723,988 | I_kwDOQb6kR874trcU | 57,860 | https://github.com/openclaw/openclaw/issues/57860 | https://api.github.com/repos/openclaw/openclaw/issues/57860 | Custom provider auth regression: codex-lb loses apiKey in embedded runtime | ## Summary
`codex-lb` as a custom OpenAI-compatible provider (`/v1`, `api: "openai-completions"`) used to work in OpenClaw, but in current versions it fails with:
```text
No API key for provider: codex-lb
```
I investigated this in depth while integrating [Soju06/codex-lb](https://github.com/Soju06/codex-lb) into Op... | closed | completed | true | 1 | [] | [] | 2026-03-30T17:35:42Z | 2026-04-28T05:17:36Z | 2026-04-25T12:20:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | SHAREN | 6,128,858 | MDQ6VXNlcjYxMjg4NTg= | User | false |
openclaw/openclaw | 4,172,828,406 | I_kwDOQb6kR874uE72 | 57,864 | https://github.com/openclaw/openclaw/issues/57864 | https://api.github.com/repos/openclaw/openclaw/issues/57864 | Discord thread inbound worker blocks for 1800s on attachment (mp4) message | ## Bug Description
When a user sends a message **with an attachment (mp4 file, ~8.4MB)** in a Discord thread, the inbound worker hangs for 1800 seconds (the full timeout), blocking all subsequent messages in that thread — even ones with `@mention`.
## Steps to Reproduce
1. In a Discord guild channel with `requireMen... | closed | completed | true | 2 | [] | [] | 2026-03-30T17:50:45Z | 2026-04-28T05:17:35Z | 2026-04-25T12:20:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | qijie850 | 258,618,953 | U_kgDOD2o2SQ | User | false |
openclaw/openclaw | 4,172,875,247 | I_kwDOQb6kR874uQXv | 57,867 | https://github.com/openclaw/openclaw/issues/57867 | https://api.github.com/repos/openclaw/openclaw/issues/57867 | [Bug] Isolated cron sessions still cannot deliver to Telegram on v2026.3.28 | **Related:** #14743, #14266
Still reproducible on **v2026.3.28** (f9b1079) running on GCP e2-small (Ubuntu 24.04, no public IP, Cloud NAT for outbound).
## Setup
10 cron jobs with `sessionTarget: "isolated"` and `delivery: { mode: "announce", channel: "telegram", to: "<userId>" }`. Telegram channel is enabled and wo... | closed | completed | true | 1 | [] | [] | 2026-03-30T17:58:13Z | 2026-04-28T05:17:34Z | 2026-04-25T12:20:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rodion-m | 36,400,912 | MDQ6VXNlcjM2NDAwOTEy | User | false |
openclaw/openclaw | 4,172,886,065 | I_kwDOQb6kR874uTAx | 57,868 | https://github.com/openclaw/openclaw/issues/57868 | https://api.github.com/repos/openclaw/openclaw/issues/57868 | [Bug]: Tool call XML leaking into Discord messages | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
**Version:** OpenClaw 2026.3.28 (f9b1079)
**Channel:** Discord
**Config:** streaming: off, streamMode: block
**Issue:** When the agent uses internal tools (read, exec, apply_patch, etc.), the raw XML tool_call... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-03-30T17:59:48Z | 2026-04-28T05:17:33Z | 2026-04-25T12:20:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | SomeRandomGuyBDS | 272,296,710 | U_kgDOEDrrBg | User | false |
openclaw/openclaw | 4,173,055,060 | I_kwDOQb6kR874u8RU | 57,875 | https://github.com/openclaw/openclaw/issues/57875 | https://api.github.com/repos/openclaw/openclaw/issues/57875 | Discord: paragraphs split into separate messages mid-sentence with blockStreaming enabled | ## Summary
When `blockStreamingDefault` is `"on"` (or `channels.discord.streaming: "block"`), paragraphs in a single assistant response are visually split across separate Discord messages, with line breaks appearing mid-sentence. The full response is well under the 2,000-character limit.
## Environment
- **OpenClaw ... | closed | completed | true | 1 | [] | [] | 2026-03-30T18:24:15Z | 2026-04-28T05:17:33Z | 2026-04-25T12:20:43Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jakepresent | 31,285,242 | MDQ6VXNlcjMxMjg1MjQy | User | false |
openclaw/openclaw | 4,173,065,918 | I_kwDOQb6kR874u-6- | 57,876 | https://github.com/openclaw/openclaw/issues/57876 | https://api.github.com/repos/openclaw/openclaw/issues/57876 | Memory flush: read tool called with empty path causes crash |
## Bug Description
During pre-compaction memory flush, the `read` tool is called with an empty/null `path` parameter, causing:
1. `ENOENT: no such file or directory` error
2. Extended blocking (16+ seconds) while the system retries
## Steps to Reproduce
1. Have memory/ directory files missing or empty
2. System tri... | closed | completed | true | 1 | [] | [] | 2026-03-30T18:25:34Z | 2026-04-28T05:17:32Z | 2026-04-25T12:20:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | JudgeAllen | 4,287,323 | MDQ6VXNlcjQyODczMjM= | User | false |
openclaw/openclaw | 4,173,135,310 | I_kwDOQb6kR874vP3O | 57,880 | https://github.com/openclaw/openclaw/issues/57880 | https://api.github.com/repos/openclaw/openclaw/issues/57880 | Retry loop duplicates user message hundreds of times in context window on rate limit | ## Bug
When the upstream provider (Anthropic Claude) rate-limits a request, OpenClaw retries the message delivery — but each retry re-appends the user's inbound message to the session context. This results in the same user message appearing 500+ times in the agent's context window.
### Expected behavior
- After N fai... | closed | completed | true | 2 | [] | [] | 2026-03-30T18:35:36Z | 2026-04-28T05:17:31Z | 2026-04-25T12:20:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KattiTronics | 258,760,875 | U_kgDOD2xgqw | User | false |
openclaw/openclaw | 4,173,326,825 | I_kwDOQb6kR874v-np | 57,888 | https://github.com/openclaw/openclaw/issues/57888 | https://api.github.com/repos/openclaw/openclaw/issues/57888 | [Bug]: openclaw configure writes Telegram allowFrom as strings, but DM allowlist matching only works with numeric IDs | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
### Summary
`openclaw configure` wrote my Telegram DM allowlist entry as a quoted string:
```json
"allowFrom": ["1234567890"]
```
With that config, the bot did not respond to direct messages from the allowlis... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-30T19:03:48Z | 2026-04-28T05:17:30Z | 2026-04-25T12:21:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | physi | 8,106,558 | MDQ6VXNlcjgxMDY1NTg= | User | false |
openclaw/openclaw | 4,173,424,136 | I_kwDOQb6kR874wWYI | 57,892 | https://github.com/openclaw/openclaw/issues/57892 | https://api.github.com/repos/openclaw/openclaw/issues/57892 | iMessage bridge drops inbound messages during high session lane activity | ## Bug Description
Inbound iMessage messages are silently dropped when the session lane is processing rapid-fire events (e.g., inter-session messages from sub-agents).
## Steps to Reproduce
1. Have an active iMessage DM session with the main agent
2. Trigger multiple rapid inter-session messages (e.g., via `sessions... | closed | completed | true | 1 | [] | [] | 2026-03-30T19:18:06Z | 2026-04-28T05:17:30Z | 2026-04-25T12:21:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Orionation | 260,579,663 | U_kgDOD4ghTw | User | false |
openclaw/openclaw | 4,173,424,243 | I_kwDOQb6kR874wWZz | 57,893 | https://github.com/openclaw/openclaw/issues/57893 | https://api.github.com/repos/openclaw/openclaw/issues/57893 | [Bug] Feishu slash commands not working in multi-account setup | ## Description
Feishu slash commands (/status, /new, /help, etc.) are not working in multi-account setup. The commands are detected but not delivered to the agent (delivered=false).
## Environment
- OpenClaw version: latest (2026.3.x)
- Channel: Feishu (websocket mode)
- Configuration: Multi-account setup with accou... | closed | completed | true | 1 | [] | [] | 2026-03-30T19:18:07Z | 2026-04-28T05:17:29Z | 2026-04-25T12:21:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | happy5318 | 34,227,488 | MDQ6VXNlcjM0MjI3NDg4 | User | false |
openclaw/openclaw | 4,173,451,527 | I_kwDOQb6kR874wdEH | 57,896 | https://github.com/openclaw/openclaw/issues/57896 | https://api.github.com/repos/openclaw/openclaw/issues/57896 | [Bug]: Gateway restart leaves stale exec config — onRestart passes activate:false to activateRuntimeSecrets | **Description:**
Full gateway restart does not refresh the runtimeConfigSnapshot. The `onRestart` handler passes `activate: false` to `activateRuntimeSecrets`, meaning exec config (safeBins, allowlist, etc.) loaded at startup is stale after a restart. Discord session lanes hold the stale config.
**Impact:**
- After a ... | closed | not_planned | true | 3 | [] | [] | 2026-03-30T19:21:41Z | 2026-04-28T05:17:28Z | 2026-04-25T12:23:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ion-ctrl | 266,816,281 | U_kgDOD-dLGQ | User | false |
openclaw/openclaw | 4,173,522,200 | I_kwDOQb6kR874wuUY | 57,900 | https://github.com/openclaw/openclaw/issues/57900 | https://api.github.com/repos/openclaw/openclaw/issues/57900 | Subagent runs don't use model fallback chain on 429/rate-limit errors | ## Bug Description
When a subagent is spawned via `sessions_spawn`, the configured fallback chain is not used when the primary model returns a 429 rate-limit error. The run fails with `LiveSessionModelSwitchError` instead of falling back.
## Expected Behavior
429 on primary → exhaust profile rotation → move to next ... | closed | completed | true | 1 | [] | [] | 2026-03-30T19:32:15Z | 2026-04-28T05:17:28Z | 2026-04-25T12:24:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Adventures-in-reality | 212,449,704 | U_kgDODKm5qA | User | false |
openclaw/openclaw | 4,173,590,466 | I_kwDOQb6kR874w-_C | 57,905 | https://github.com/openclaw/openclaw/issues/57905 | https://api.github.com/repos/openclaw/openclaw/issues/57905 | Gateway enters infinite model-switch loop when all auth profiles fail | ## Problem
When all auth profiles enter cooldown (e.g., Anthropic OAuth token rejected + no xAI key), the gateway enters an infinite model-switch loop cycling every ~1 second. The gateway becomes completely unresponsive and cannot process any messages. The loop persists across restarts because session state is recreate... | closed | completed | true | 1 | [] | [] | 2026-03-30T19:42:50Z | 2026-04-28T05:17:27Z | 2026-04-25T12:24:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | klausbot98-bot | 262,215,816 | U_kgDOD6EYiA | User | false |
openclaw/openclaw | 4,173,591,549 | I_kwDOQb6kR874w_P9 | 57,906 | https://github.com/openclaw/openclaw/issues/57906 | https://api.github.com/repos/openclaw/openclaw/issues/57906 | Model fallback retries primary too aggressively before moving to next fallback | ## Problem
When the primary model (e.g., Opus) returns overloaded_error, the gateway retries it for 30-60 seconds before attempting the first fallback (Sonnet). Even then, if Sonnet is also overloaded, GPT-5.4 as the third fallback is rarely reached.
## Expected behavior
After 1-2 failed attempts on primary (within ~5... | closed | completed | true | 2 | [] | [] | 2026-03-30T19:43:02Z | 2026-04-28T05:17:26Z | 2026-04-25T12:24:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | klausbot98-bot | 262,215,816 | U_kgDOD6EYiA | User | false |
openclaw/openclaw | 4,173,592,752 | I_kwDOQb6kR874w_iw | 57,907 | https://github.com/openclaw/openclaw/issues/57907 | https://api.github.com/repos/openclaw/openclaw/issues/57907 | openclaw configure --section model leaves stale claude-cli/ prefix in model config | ## Problem
Running `openclaw configure --section model` and choosing 'Anthropic Claude CLI' sets the model to `claude-cli/claude-opus-4-6` which is not a valid model ID. This results in `Unknown model: claude-cli/claude-opus-4-6 (model_not_found)` errors.
The fallback models also get set with the `claude-cli/` prefix,... | closed | not_planned | true | 1 | [] | [] | 2026-03-30T19:43:13Z | 2026-04-28T05:17:25Z | 2026-04-25T12:24:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | klausbot98-bot | 262,215,816 | U_kgDOD6EYiA | User | false |
openclaw/openclaw | 4,173,605,377 | I_kwDOQb6kR874xCoB | 57,908 | https://github.com/openclaw/openclaw/issues/57908 | https://api.github.com/repos/openclaw/openclaw/issues/57908 | [Feature]: Allow prompt-level selection of LLM service provider | ### Summary
Allow use of prompt-level selection of LLM provider by way of some syntax similar to
llmservers.yaml
ollama:
servers:
- url: "http://127.0.0.1:11434"
name: "Server1"
priority: 1
primary: true
- url: "http://192.168.1.99:11434"
name: "Server2"
priority: 2
backup... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-30T19:45:20Z | 2026-04-28T05:17:25Z | 2026-04-25T12:24:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fmotta | 877,390 | MDQ6VXNlcjg3NzM5MA== | User | false |
openclaw/openclaw | 4,173,732,947 | I_kwDOQb6kR874xhxT | 57,916 | https://github.com/openclaw/openclaw/issues/57916 | https://api.github.com/repos/openclaw/openclaw/issues/57916 | Subagent announce: swap dispatch order to queue-first when parent session is busy | ## Summary
Subagent announce completion messages use `direct-primary` dispatch (via `callGateway`) as the first attempt for `expectsCompletionMessage: true`. When the parent session has an active run, the direct call queues into the session's command lane and blocks until the lane is free — often timing out.
**Propos... | closed | completed | true | 1 | [] | [] | 2026-03-30T20:05:35Z | 2026-04-28T05:17:24Z | 2026-04-25T12:24:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | arcimun | 9,164,851 | MDQ6VXNlcjkxNjQ4NTE= | User | false |
openclaw/openclaw | 4,328,075,526 | I_kwDOQb6kR88AAAABAfkxBg | 71,575 | https://github.com/openclaw/openclaw/issues/71575 | https://api.github.com/repos/openclaw/openclaw/issues/71575 | Telegram channel renders intermediate tool call steps as visible messages | ## Description
When an agent is running in a Telegram-bound session and executes tool calls (e.g. `exec`, `memory_search`, etc.), the intermediate "Working…" status updates and raw tool output are rendered as visible messages in the Telegram chat. Only the final assistant reply should be delivered to the user.
## Exp... | closed | completed | false | 4 | [] | [] | 2026-04-25T12:24:43Z | 2026-05-04T06:42:32Z | 2026-05-04T06:42:32Z | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | adamwdear | 258,995,361 | U_kgDOD2_0oQ | User | false |
openclaw/openclaw | 4,173,735,749 | I_kwDOQb6kR874xidF | 57,917 | https://github.com/openclaw/openclaw/issues/57917 | https://api.github.com/repos/openclaw/openclaw/issues/57917 | Control UI model switch misparses aliases like gpt-5.4 as anthropic/* | ## Bug Description
In the Control UI, manually switching models with a short alias/name like `gpt-5.4` can get misparsed as if it belongs to the current/default provider. Example error shown in UI:
`Failed to set model: GatewayRequestError: model not allowed: anthropic/gpt 5.4`
The backend model switch works correct... | closed | completed | true | 1 | [] | [] | 2026-03-30T20:05:59Z | 2026-04-28T05:17:23Z | 2026-04-25T12:24:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Adventures-in-reality | 212,449,704 | U_kgDODKm5qA | User | false |
openclaw/openclaw | 4,173,739,436 | I_kwDOQb6kR874xjWs | 57,918 | https://github.com/openclaw/openclaw/issues/57918 | https://api.github.com/repos/openclaw/openclaw/issues/57918 | [Bug]: Announce delivery constructs malformed recipient "group:<chatId>" for Telegram forum topic sessions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When sessions_send completes an async (fire-and-forget) delegation between agents in a Telegram forum topic session, the announce delivery step constructs the recipient as "group:-10...2" instead of parsing the... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T20:06:32Z | 2026-04-28T05:17:23Z | 2026-04-25T12:24:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dLcch | 170,981,687 | U_kgDOCjD5Nw | User | false |
openclaw/openclaw | 4,173,855,148 | I_kwDOQb6kR874x_ms | 57,925 | https://github.com/openclaw/openclaw/issues/57925 | https://api.github.com/repos/openclaw/openclaw/issues/57925 | v2026.3.28: Discord extension silently skipped during gateway startup — zero channels load despite valid config | # Bug Report: v2026.3.28 Discord channels silently fail to load on macOS
## Title
v2026.3.28: Discord extension silently skipped during gateway startup — zero channels load despite valid config
## Environment
- **OpenClaw:** v2026.3.28 (f9b1079)
- **OS:** macOS (Darwin 25.2.0, arm64, Mac Mini)
- **Node:** v25.8.2
- *... | closed | completed | true | 1 | [] | [] | 2026-03-30T20:24:57Z | 2026-04-28T05:17:22Z | 2026-04-25T12:25:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | samanthatwombly | 271,669,637 | U_kgDOEDFZhQ | User | false |
openclaw/openclaw | 4,173,890,772 | I_kwDOQb6kR874yITU | 57,928 | https://github.com/openclaw/openclaw/issues/57928 | https://api.github.com/repos/openclaw/openclaw/issues/57928 | session.tool / session.message events don't carry originating sessionKey for subagent sessions | ## Summary
When a subagent (spawned via `sessions_spawn`) makes tool calls, the resulting `session.tool` and `session.message` events broadcast over the gateway WebSocket do not include the child session's `sessionKey`. They arrive attributed to the parent session (`agent:main:main`).
This makes it impossible for ope... | closed | completed | true | 1 | [] | [] | 2026-03-30T20:31:22Z | 2026-04-28T05:17:21Z | 2026-04-25T12:25:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | joechristie1020 | 262,193,448 | U_kgDOD6DBKA | User | false |
openclaw/openclaw | 4,173,904,267 | I_kwDOQb6kR874yLmL | 57,930 | https://github.com/openclaw/openclaw/issues/57930 | https://api.github.com/repos/openclaw/openclaw/issues/57930 | openai-codex-responses provider sends oversized `instructions` field, gets 400 Bad Request | # openai-codex-responses provider sends oversized `instructions` field, gets 400 Bad Request
## Summary
The `openai-codex` provider (targeting `chatgpt.com/backend-api/codex/responses`) does not cap the `instructions` field before sending. When workspace bootstrap files + system prompt + tool schemas exceed ~32 KiB, ... | closed | completed | true | 1 | [] | [] | 2026-03-30T20:33:42Z | 2026-04-28T05:17:20Z | 2026-04-25T12:25:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | barnaby-coder | 257,451,284 | U_kgDOD1hlFA | User | false |
openclaw/openclaw | 4,174,009,880 | I_kwDOQb6kR874ylYY | 57,939 | https://github.com/openclaw/openclaw/issues/57939 | https://api.github.com/repos/openclaw/openclaw/issues/57939 | [Bug]: Gateway default model not updated after `openclaw models set` + service restart | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
After setting a new default model with `openclaw models set` and performing a full gateway stop/start, new Telegram sessions continue to report the old default model rather than the newly configured one.
### S... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-03-30T20:52:04Z | 2026-04-28T05:17:20Z | 2026-04-25T12:25:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | shuntera | 11,050,796 | MDQ6VXNlcjExMDUwNzk2 | User | false |
openclaw/openclaw | 4,174,040,774 | I_kwDOQb6kR874ys7G | 57,941 | https://github.com/openclaw/openclaw/issues/57941 | https://api.github.com/repos/openclaw/openclaw/issues/57941 | Plugin hook: support per-message model switching in persistent sessions | ## Feature Request
### Problem
The `before_model_resolve` plugin hook can only influence the model at session start. When a session is already running with model A and the hook returns a different model for a subsequent message, OpenClaw throws a `LiveSessionModelSwitchError` instead of honoring the override.
**Erro... | closed | completed | true | 1 | [] | [] | 2026-03-30T20:57:22Z | 2026-04-28T05:17:19Z | 2026-04-25T12:25:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | saibot1984 | 272,331,288 | U_kgDOEDtyGA | User | false |
openclaw/openclaw | 4,174,144,966 | I_kwDOQb6kR874zGXG | 57,945 | https://github.com/openclaw/openclaw/issues/57945 | https://api.github.com/repos/openclaw/openclaw/issues/57945 | Bug: Isolated agentTurn cron jobs fail - docs/reference/templates not bundled in npm package | ## Bug: Isolated agentTurn cron jobs fail with "Missing workspace template: AGENTS.md"
**Version:** 2026.3.28 (f9b1079)
**OS:** Windows 11 (x64)
**Severity:** Critical — blocks entire cron fleet
### Summary
All isolated `agentTurn` cron jobs fail at launch with the following error before any payload runs:
```
Error... | closed | completed | true | 1 | [] | [] | 2026-03-30T21:19:03Z | 2026-04-28T05:17:18Z | 2026-04-25T12:25:43Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jvpflum | 13,372,521 | MDQ6VXNlcjEzMzcyNTIx | User | false |
openclaw/openclaw | 4,174,214,853 | I_kwDOQb6kR874zXbF | 57,949 | https://github.com/openclaw/openclaw/issues/57949 | https://api.github.com/repos/openclaw/openclaw/issues/57949 | ClawHub install for @openguardrails/moltguard fails with missing zip in /tmp | ## Bug Description
Installing the ClawHub plugin `@openguardrails/moltguard` fails after successful package resolution.
The CLI resolves metadata correctly, shows compatibility, and recognizes the package as a community code-plugin, but then fails trying to open the downloaded zip payload from `/tmp`.
## Command
``... | closed | completed | true | 1 | [] | [] | 2026-03-30T21:34:41Z | 2026-04-28T05:17:17Z | 2026-04-25T12:25:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Adventures-in-reality | 212,449,704 | U_kgDODKm5qA | User | false |
openclaw/openclaw | 4,174,384,334 | I_kwDOQb6kR8740AzO | 57,960 | https://github.com/openclaw/openclaw/issues/57960 | https://api.github.com/repos/openclaw/openclaw/issues/57960 | Discord threads should inherit parent channel agent binding and exec permissions | ## Problem
Discord threads do not inherit the parent channel's agent binding. A thread created in a channel bound to agent `dev` falls back to the default agent, which has different (more restrictive) exec permissions. This forces operators to either:
- Add explicit bindings for every new thread ID
- Use a guild-level... | closed | completed | true | 1 | [] | [] | 2026-03-30T22:12:12Z | 2026-04-28T05:17:16Z | 2026-04-25T12:26:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | leosmigel | 7,751,761 | MDQ6VXNlcjc3NTE3NjE= | User | false |
openclaw/openclaw | 4,174,404,516 | I_kwDOQb6kR8740Fuk | 57,961 | https://github.com/openclaw/openclaw/issues/57961 | https://api.github.com/repos/openclaw/openclaw/issues/57961 | [Bug]: Live session model switch detector blocks programmatic fallback during rate limits | ## Bug Report
### Summary
When a running session hits a rate limit (HTTP 429) and the fallback chain attempts to rotate to a different provider, the "live session model switch" detector fires and **cancels the fallback attempt**, forcing the session back to the rate-limited provider. This creates a loop where no fall... | closed | completed | true | 1 | [] | [] | 2026-03-30T22:16:50Z | 2026-04-28T05:17:16Z | 2026-04-25T12:26:12Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | LeslieCBarry | 195,984,913 | U_kgDOC65-EQ | User | false |
openclaw/openclaw | 4,174,446,319 | I_kwDOQb6kR8740P7v | 57,965 | https://github.com/openclaw/openclaw/issues/57965 | https://api.github.com/repos/openclaw/openclaw/issues/57965 | Sub-agent exec output not returned to parent session | ## Bug Description
When a persistent agent spawns a sub-agent (`runtime=subagent`) that runs exec commands, the exec runs successfully (confirmed via token counts) but the stdout/stderr output is NOT returned to the parent session. The parent session only sees the tool call metadata, not the actual command output.
##... | closed | completed | true | 1 | [] | [] | 2026-03-30T22:27:31Z | 2026-04-28T05:17:15Z | 2026-04-25T12:26:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | DSchexy | 262,326,558 | U_kgDOD6LJHg | User | false |
openclaw/openclaw | 4,174,466,086 | I_kwDOQb6kR8740Uwm | 57,969 | https://github.com/openclaw/openclaw/issues/57969 | https://api.github.com/repos/openclaw/openclaw/issues/57969 | [Bug]: Zapier MCP via @ownid/mcp-remote fails in OpenClaw bundle MCP with MCP error -32001: Request timed out | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
## Summary
I'm trying to use Zapier MCP in OpenClaw via a stdio bridge subprocess:
```json
"mcp": {
"servers": {
"zapier": {
"command": "npx",
"args": [
"-y",
"@ownid/mcp-remo... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-30T22:31:33Z | 2026-04-28T05:17:14Z | 2026-04-25T12:26:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cnsandbox | 231,573,167 | U_kgDODc2Grw | User | false |
openclaw/openclaw | 4,174,522,823 | I_kwDOQb6kR8740inH | 57,973 | https://github.com/openclaw/openclaw/issues/57973 | https://api.github.com/repos/openclaw/openclaw/issues/57973 | [Bug]: edit tool 100% failure in 2026.3.30 — CLAUDE_PARAM_GROUPS.edit validation rejects edits[] array format | ### Bug type
Regression (worked before, now fails)
### Summary
The `edit` tool fails with `Missing required parameters: oldText alias, newText alias` on every invocation since gateway restart on OpenClaw 2026.3.30.
### Root cause analysis
`CLAUDE_PARAM_GROUPS.edit` in `pi-tools.params.ts` expects `oldText` and `ne... | closed | completed | true | 4 | [] | [] | 2026-03-30T22:46:18Z | 2026-04-28T05:17:13Z | 2026-04-25T12:26:33Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kagura-agent | 268,167,063 | U_kgDOD_vnlw | User | false |
openclaw/openclaw | 4,174,545,114 | I_kwDOQb6kR8740oDa | 57,974 | https://github.com/openclaw/openclaw/issues/57974 | https://api.github.com/repos/openclaw/openclaw/issues/57974 | control-ui sessions.list polling saturates gateway with many conversations | ## Bug
**Version:** v2026.3.28
### Problem
The `openclaw-control-ui` webchat client polls `sessions.list` every ~5-7 seconds per connected tab. With a large LCM database (~5,000 conversations, 163K messages), each `sessions.list` call takes **~4.8 seconds** to resolve.
Two open tabs (overview + chat) produce ~24 ca... | closed | not_planned | true | 1 | [] | [] | 2026-03-30T22:53:51Z | 2026-04-28T05:17:13Z | 2026-04-25T12:26:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Veda-openclaw | 267,342,076 | U_kgDOD-9Q_A | User | false |
openclaw/openclaw | 4,174,569,986 | I_kwDOQb6kR8740uIC | 57,977 | https://github.com/openclaw/openclaw/issues/57977 | https://api.github.com/repos/openclaw/openclaw/issues/57977 | [Feature]: Continuous sessions across interfaces. | ### Summary
I would like to be able to start a conversation in, say, the web interface, then pick it up in the SMS interface.
### Problem to solve
I often move from location to location, and need to continue a conversation. Being able to do this rapidly would be a great help.
### Proposed solution
Automating this... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-03-30T23:01:40Z | 2026-04-28T05:17:12Z | 2026-04-25T12:26:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chillyjim | 17,027,445 | MDQ6VXNlcjE3MDI3NDQ1 | User | false |
openclaw/openclaw | 4,174,647,410 | I_kwDOQb6kR8741BBy | 57,988 | https://github.com/openclaw/openclaw/issues/57988 | https://api.github.com/repos/openclaw/openclaw/issues/57988 | Bug: Heartbeat read tool called without path | ## Description
The heartbeat scheduler is calling the read tool without passing the `path` parameter correctly.
## Error Log
```
[agent/embedded] read tool called without path: toolCallId=call_391c245501c54387b19a9c50 argsType=object
```
## Details
-**Version:** OpenClaw 2026.3.28 (f9b1079)
-**Frequency:** Every ~... | closed | completed | true | 2 | [] | [] | 2026-03-30T23:24:04Z | 2026-04-28T05:17:11Z | 2026-04-25T12:26:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hamodae | 883,688 | MDQ6VXNlcjg4MzY4OA== | User | false |
openclaw/openclaw | 4,174,742,651 | I_kwDOQb6kR8741YR7 | 57,997 | https://github.com/openclaw/openclaw/issues/57997 | https://api.github.com/repos/openclaw/openclaw/issues/57997 | Cron jobs ignore payload.model and run on agent default model (Opus instead of GPT-5.4) | ## Bug Description
Cron jobs that specify `payload.model` in their configuration are **ignoring that field** and running on the agent's default model instead.
## Expected Behavior
A cron job configured with:
```json
"payload": {
"kind": "agentTurn",
"model": "openai-codex/gpt-5.4",
...
}
```
Should execute us... | closed | completed | true | 1 | [] | [] | 2026-03-30T23:54:48Z | 2026-04-28T05:17:10Z | 2026-04-25T12:27:02Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | beto-sudo | 229,640,506 | U_kgDODbAJOg | User | false |
openclaw/openclaw | 4,174,770,577 | I_kwDOQb6kR8741fGR | 57,999 | https://github.com/openclaw/openclaw/issues/57999 | https://api.github.com/repos/openclaw/openclaw/issues/57999 | Subagent deliveryContext loses to/threadId/accountId when spawned from group topics | ## Bug Description
When a subagent is spawned via `sessions_spawn` from a **group topic session** (e.g. Telegram forum topic), the subagent's `deliveryContext` only inherits `{"channel": "telegram"}` — it **loses** the `to`, `accountId`, and `threadId` fields.
This means the completion announce has nowhere to be deli... | closed | completed | true | 1 | [] | [] | 2026-03-31T00:03:19Z | 2026-04-28T05:17:10Z | 2026-04-25T12:27:09Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ali3412-lgtm | 229,473,436 | U_kgDODa18nA | User | false |
openclaw/openclaw | 4,174,816,365 | I_kwDOQb6kR8741qRt | 58,004 | https://github.com/openclaw/openclaw/issues/58004 | https://api.github.com/repos/openclaw/openclaw/issues/58004 | Feature: Cross-channel session binding (unified session across TUI + Telegram/Discord) | ## Problem
Currently each surface (TUI/webchat, Telegram DM, Discord) gets its own session key. Context doesn't carry when switching between them — you have to re-establish state every time you move from mobile Telegram to the local TUI or vice versa.
## Proposed Solution
Allow mapping multiple channel/sender identi... | closed | completed | true | 1 | [] | [] | 2026-03-31T00:17:28Z | 2026-04-28T05:17:09Z | 2026-04-25T12:27:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ryanpreyes-stack | 266,851,092 | U_kgDOD-fTFA | User | false |
openclaw/openclaw | 4,174,851,981 | I_kwDOQb6kR8741y-N | 58,009 | https://github.com/openclaw/openclaw/issues/58009 | https://api.github.com/repos/openclaw/openclaw/issues/58009 | docs/config: plugins.allow allowlist silently blocks all stock channel plugins (Discord, Signal, etc.) when set | ## Summary
Setting `plugins.allow` in `openclaw.json` to a non-empty array silently blocks **all** stock/bundled channel plugins — including Discord, Signal, and others — unless they are explicitly listed. There is no warning during `openclaw doctor` or at gateway startup to indicate that channels have been dropped.
... | closed | completed | true | 2 | [] | [] | 2026-03-31T00:29:39Z | 2026-04-28T05:17:08Z | 2026-04-25T12:27:40Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KR-Python | 48,324,619 | MDQ6VXNlcjQ4MzI0NjE5 | User | false |
openclaw/openclaw | 4,174,979,396 | I_kwDOQb6kR8742SFE | 58,031 | https://github.com/openclaw/openclaw/issues/58031 | https://api.github.com/repos/openclaw/openclaw/issues/58031 | feat(workspace): Increase AGENTS.md/SOUL.md context file size limits (4KB/3KB too restrictive for complex agent configurations) | ## Problem
When operating autonomous research agents with complex pipelines, the 4KB limit for `AGENTS.md` and 3KB limit for `SOUL.md` become restrictive.
**Current state (monitored over multiple sessions):**
- `AGENTS.md`: 3,934 / 4,000 bytes (98%) — cannot add any new operational guidance
- `SOUL.md`: 2,853 / 3,000... | closed | completed | true | 1 | [] | [] | 2026-03-31T01:07:49Z | 2026-04-28T05:17:07Z | 2026-04-25T12:27:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gasvn | 20,515,144 | MDQ6VXNlcjIwNTE1MTQ0 | User | false |
openclaw/openclaw | 4,174,983,356 | I_kwDOQb6kR8742TC8 | 58,033 | https://github.com/openclaw/openclaw/issues/58033 | https://api.github.com/repos/openclaw/openclaw/issues/58033 | Feature Request: Silent Heartbeat Mode | '## Description
Add a silent heartbeat mode that runs health checks silently without sending messages to the chat frontend. This would prevent the heartbeat messages from disrupting UI elements like dialogs in the webchat interface.
## Problem
Currently, heartbeat messages (e.g., HEARTBEAT_OK) are sent to the chat f... | closed | completed | true | 1 | [] | [] | 2026-03-31T01:09:02Z | 2026-04-28T05:17:07Z | 2026-04-25T12:27:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | limuran117-coder | 266,618,023 | U_kgDOD-REpw | User | false |
openclaw/openclaw | 4,174,993,765 | I_kwDOQb6kR8742Vll | 58,039 | https://github.com/openclaw/openclaw/issues/58039 | https://api.github.com/repos/openclaw/openclaw/issues/58039 | /model can become unusable when the current provider is unavailable, and configured fallbacks do not recover the session | ## Summary
When the currently selected model provider becomes unavailable, `/model` may also become unusable, and configured `fallbacks` do not recover the session.
This makes the session effectively stuck on a broken provider, which feels incorrect because `/model` is a control action and should still allow switchin... | closed | completed | true | 2 | [] | [] | 2026-03-31T01:12:36Z | 2026-04-28T05:17:06Z | 2026-04-25T12:28:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mp1802174 | 71,064,038 | MDQ6VXNlcjcxMDY0MDM4 | User | false |
openclaw/openclaw | 4,174,995,336 | I_kwDOQb6kR8742V-I | 58,041 | https://github.com/openclaw/openclaw/issues/58041 | https://api.github.com/repos/openclaw/openclaw/issues/58041 | Auto-update is not atomic: config/plugin version mismatch causes repeated crash loops | ## Summary
Auto-update consistently causes gateway crash loops because it is not atomic — config or plugin manifests get updated to require a newer version while the binary remains at the old version (or vice versa). This has caused **3 separate overnight outages in 6 days** on our setup.
## Environment
- macOS 15.x... | closed | completed | true | 2 | [] | [] | 2026-03-31T01:13:06Z | 2026-04-28T05:17:05Z | 2026-04-25T12:28:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | allanjeng | 14,968,942 | MDQ6VXNlcjE0OTY4OTQy | User | false |
openclaw/openclaw | 4,175,086,454 | I_kwDOQb6kR8742sN2 | 58,045 | https://github.com/openclaw/openclaw/issues/58045 | https://api.github.com/repos/openclaw/openclaw/issues/58045 | LiveSessionModelSwitchError prevents cross-provider model fallback | ## Summary
When the primary model (e.g. `anthropic/claude-sonnet-4-5`) returns a 429 rate limit error, the gateway correctly decides to fall back to the next configured model (e.g. `google/gemini-2.5-flash`). However, the embedded run's `resolvePersistedLiveSelection()` detects a "model switch" and throws `LiveSession... | closed | completed | true | 1 | [] | [] | 2026-03-31T01:43:09Z | 2026-04-28T05:17:04Z | 2026-04-25T12:28:16Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mids-neo | 117,532,760 | U_kgDOBwFoWA | User | false |
openclaw/openclaw | 4,175,088,500 | I_kwDOQb6kR8742st0 | 58,046 | https://github.com/openclaw/openclaw/issues/58046 | https://api.github.com/repos/openclaw/openclaw/issues/58046 | Foundry plugin enters infinite reload loop, crashes gateway | ## Description
The `foundry-openclaw` plugin enters an infinite reload loop that crashes the OpenClaw gateway. The plugin re-registers itself multiple times per second (up to 1,448 reload cycles in a single day), eventually exhausting memory and killing the process without a clean exit log.
## Environment
- OpenClaw... | closed | completed | true | 1 | [] | [] | 2026-03-31T01:43:46Z | 2026-04-28T05:17:04Z | 2026-04-25T12:28:23Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dbigman | 23,329,506 | MDQ6VXNlcjIzMzI5NTA2 | User | false |
openclaw/openclaw | 4,175,173,547 | I_kwDOQb6kR8743Ber | 58,060 | https://github.com/openclaw/openclaw/issues/58060 | https://api.github.com/repos/openclaw/openclaw/issues/58060 | [Feature]: Add Ollama as an interactive setup assistant option during OpenClaw installation | ### Summary
Add Ollama as an interactive setup assistant option during OpenClaw installation
### Problem to solve
Non-technical users struggle with OpenClaw installation because they need to understand technical terms (model provider, API key, Gateway bind, etc.). Current `openclaw onboard` still requires users to m... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-31T02:07:28Z | 2026-04-28T05:17:03Z | 2026-04-25T12:28:30Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kuniakil | 192,518,029 | U_kgDOC3mXjQ | User | false |
openclaw/openclaw | 4,175,252,061 | I_kwDOQb6kR8743Upd | 58,069 | https://github.com/openclaw/openclaw/issues/58069 | https://api.github.com/repos/openclaw/openclaw/issues/58069 | [Bug]: Live session model switch overrides active fallback during rate-limit cooldown, creating infinite retry loop | ## Description
When an Anthropic model hits a rate limit (HTTP 429), the fallback engine correctly selects the next model in the fallback chain (e.g. `openai/gpt-5.4`). However, the **"live session model switch"** feature immediately overrides this fallback decision, resetting the model back to the rate-limited primar... | closed | completed | true | 2 | [] | [] | 2026-03-31T02:33:42Z | 2026-04-28T05:17:02Z | 2026-04-25T12:28:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | taswest81 | 222,143,630 | U_kgDODT2kjg | User | false |
openclaw/openclaw | 4,175,274,671 | I_kwDOQb6kR8743aKv | 58,076 | https://github.com/openclaw/openclaw/issues/58076 | https://api.github.com/repos/openclaw/openclaw/issues/58076 | [Feature Request] Allow spawning named sub-agents via sessions_spawn for configured agents | ## Problem
Currently, `sessions_spawn` with `runtime=\"subagent\"` does not allow targeting named agents (e.g. front-02, flow, spec) even when those agents are properly configured in `agents.list`. The API returns:
```
error: \"agentId is not allowed for sessions_spawn (allowed: none)\"
```
There is no configuration... | closed | completed | true | 1 | [] | [] | 2026-03-31T02:40:39Z | 2026-04-28T05:17:01Z | 2026-04-25T12:28:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jackycxl | 200,453,816 | U_kgDOC_KuuA | User | false |
openclaw/openclaw | 4,175,323,092 | I_kwDOQb6kR8743l_U | 58,081 | https://github.com/openclaw/openclaw/issues/58081 | https://api.github.com/repos/openclaw/openclaw/issues/58081 | fix(coding): gh issue view uses invalid `reactions` field — should be `reactionGroups` | ## Problem
`coding_start_workflow` consistently fails because the coding plugin calls:
```
gh issue view --json number,title,body,url,labels,comments,reactions,assignees,createdAt
```
But `reactions` is not a valid field for `gh issue view`. The correct field is `reactionGroups`.
Error from logs (2026-03-30, repeat... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T02:55:35Z | 2026-04-28T05:17:00Z | 2026-04-25T12:28:51Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nqhoag | 23,328,947 | MDQ6VXNlcjIzMzI4OTQ3 | User | false |
openclaw/openclaw | 4,328,081,052 | I_kwDOQb6kR88AAAABAflGnA | 71,576 | https://github.com/openclaw/openclaw/issues/71576 | https://api.github.com/repos/openclaw/openclaw/issues/71576 | [Feature]: Make the macOS updater/release pipeline fully Sparkle-native and atomic | ## Summary
Make the macOS companion app's release/update path fully standardized around `sparkle-project/Sparkle`, with one atomic publishing flow so the appcast can never advertise a missing or mismatched asset.
## Problem to solve
OpenClaw already uses Sparkle in the macOS app (`apps/macos/Package.swift`) and the ... | closed | not_planned | true | 0 | [] | [] | 2026-04-25T12:27:14Z | 2026-04-28T05:16:59Z | 2026-04-25T12:28:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nsd97 | 5,872,983 | MDQ6VXNlcjU4NzI5ODM= | User | false |
openclaw/openclaw | 4,175,323,704 | I_kwDOQb6kR8743mI4 | 58,082 | https://github.com/openclaw/openclaw/issues/58082 | https://api.github.com/repos/openclaw/openclaw/issues/58082 | fix(coding): coding_start_workflow fails with invalid repo format "/" when repo not configured | ## Problem
When `coding_start_workflow` is called without a valid GitHub repo configured, it constructs an invalid `"/"` repo string and calls:
```
gh issue comment 1 --repo / --body "PA coding workflow started..."
```
Resulting in: `expected the "[HOST/]OWNER/REPO" format, got "/"` — repeated 10x in logs.
## Fix
... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T02:55:45Z | 2026-04-28T05:16:58Z | 2026-04-25T12:28:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nqhoag | 23,328,947 | MDQ6VXNlcjIzMzI4OTQ3 | User | false |
openclaw/openclaw | 4,175,330,276 | I_kwDOQb6kR8743nvk | 58,084 | https://github.com/openclaw/openclaw/issues/58084 | https://api.github.com/repos/openclaw/openclaw/issues/58084 | [Bug] HTTP 400 tools.N.custom.input_schema: JSON schema is invalid — blocks all tool use | ## Bug Description
After updating to OpenClaw 2026.3.28, all sub-agents (and sometimes main agent) fail with:
```
HTTP 400: tools.27.custom.input_schema: JSON schema is invalid. It must match JSON Schema draft 2020-12
```
The error occurs when any tool is invoked (e.g. `read`, file operations). The error is consiste... | closed | completed | true | 1 | [] | [] | 2026-03-31T02:57:42Z | 2026-04-28T05:16:58Z | 2026-04-25T12:29:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jackycxl | 200,453,816 | U_kgDOC_KuuA | User | false |
openclaw/openclaw | 4,175,354,817 | I_kwDOQb6kR8743tvB | 58,087 | https://github.com/openclaw/openclaw/issues/58087 | https://api.github.com/repos/openclaw/openclaw/issues/58087 | [Bug]: SecretRef-backed model provider headers sent as literal "secretref-managed" in API requests | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from v2026.3.24 to v2026.3.28, custom model provider headers configured via SecretRef are sent to the upstream API as the literal placeholder string `"secretref-managed"` instead of the resolved secret value,... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-31T03:04:39Z | 2026-04-28T05:16:57Z | 2026-04-25T12:29:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | TheWinds | 13,296,462 | MDQ6VXNlcjEzMjk2NDYy | User | false |
openclaw/openclaw | 4,175,357,454 | I_kwDOQb6kR8743uYO | 58,090 | https://github.com/openclaw/openclaw/issues/58090 | https://api.github.com/repos/openclaw/openclaw/issues/58090 | openclaw doctor wipes session.resetByType config | ## Summary
Running `openclaw doctor` resets `session.resetByType` back to defaults, removing custom idle timeout configuration. This causes group and thread sessions to unexpectedly reset based on default timing instead of the user-configured values.
## Environment
- OpenClaw latest (as of 2026-03-30)
- Channel: Tel... | closed | completed | true | 1 | [] | [] | 2026-03-31T03:05:22Z | 2026-04-28T05:16:56Z | 2026-04-25T12:29:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | drinkflav | 256,454,846 | U_kgDOD0kwvg | User | false |
openclaw/openclaw | 4,175,364,241 | I_kwDOQb6kR8743wCR | 58,091 | https://github.com/openclaw/openclaw/issues/58091 | https://api.github.com/repos/openclaw/openclaw/issues/58091 | [Feature]: add persistent todo tool and long-running workflow skill for agent tasks | ### Summary
Add a persistent `todo` tool and a long-running workflow skill so agents can plan, track, resume, and hand off multi-step work across Feishu conversations, local CLI execution, and spawned subagent/ACP sessions.
### Work type
feature:new
### Problem to solve
OpenClaw already has strong execution primitive... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T03:07:13Z | 2026-04-28T05:16:55Z | 2026-04-25T12:29:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | oricter | 39,150,429 | MDQ6VXNlcjM5MTUwNDI5 | User | false |
openclaw/openclaw | 4,175,372,546 | I_kwDOQb6kR8743yEC | 58,094 | https://github.com/openclaw/openclaw/issues/58094 | https://api.github.com/repos/openclaw/openclaw/issues/58094 | fix(pa-coding): coding_start_workflow fails to parse GitHub issue URL — owner/repo/number undefined | ## Bug
When calling `coding_start_workflow(issueUrl)`, the plugin fails to extract `owner`, `repo`, and `issue number` from the URL. The gh CLI then receives malformed arguments.
## Errors
```
# Error 1: issue number is "undefined"
coding_start_workflow failed: gh issue view invalid issue format: "undefined"
# Erro... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T03:09:28Z | 2026-04-28T05:16:54Z | 2026-04-25T12:29:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nqhoag | 23,328,947 | MDQ6VXNlcjIzMzI4OTQ3 | User | false |
openclaw/openclaw | 4,175,372,569 | I_kwDOQb6kR8743yEZ | 58,095 | https://github.com/openclaw/openclaw/issues/58095 | https://api.github.com/repos/openclaw/openclaw/issues/58095 | fix(pa-coding): coding_start_workflow fails — gh CLI reactions field invalid, should use reactionGroups | ## Bug
`coding_start_workflow` immediately fails because the gh CLI is called with `--json reactions` which is not a valid field.
## Error
```
[tools] coding_start_workflow failed: Command failed: gh issue view --json number,title,body,url,labels,comments,reactions,assignees,createdAt
Unknown JSON field: "reactions... | closed | not_planned | true | 1 | [] | [] | 2026-03-31T03:09:28Z | 2026-04-28T05:16:53Z | 2026-04-25T12:29:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nqhoag | 23,328,947 | MDQ6VXNlcjIzMzI4OTQ3 | User | false |
openclaw/openclaw | 4,175,420,208 | I_kwDOQb6kR87439sw | 58,104 | https://github.com/openclaw/openclaw/issues/58104 | https://api.github.com/repos/openclaw/openclaw/issues/58104 | [Feature]: Feishu built-in skill: feishu-sheet is missing from dist/extensions/feishu/skills/ | ### Summary
## Bug Description
The Feishu built-in skill set in `dist/extensions/feishu/skills/` includes:
- feishu-doc ✅
- feishu-drive ✅
- feishu-perm ✅
- feishu-wiki ✅
- feishu-bitable ✅
But `feishu-sheet` is missing from the built-in set.
## Expected Behavior
All Feishu core modules (doc, drive, perm, wiki, bi... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-03-31T03:22:23Z | 2026-04-28T05:16:53Z | 2026-04-25T12:29:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Hanm-Huang | 269,858,534 | U_kgDOEBW25g | User | false |
openclaw/openclaw | 4,175,453,408 | I_kwDOQb6kR8744Fzg | 58,107 | https://github.com/openclaw/openclaw/issues/58107 | https://api.github.com/repos/openclaw/openclaw/issues/58107 | [Bug]: Multiple Feishu group agents - only main reply delivered, other agents' replies silently dropped | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When multiple agents are subscribed to the same Feishu group, only the main agent's reply is delivered; other agents' replies are silently dropped despite being generated.
### Steps to reproduce
1、Configure two agents (m... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-31T03:31:16Z | 2026-04-28T05:16:52Z | 2026-04-25T12:30:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | slladmin | 58,767,080 | MDQ6VXNlcjU4NzY3MDgw | User | false |
openclaw/openclaw | 4,175,456,637 | I_kwDOQb6kR8744Gl9 | 58,108 | https://github.com/openclaw/openclaw/issues/58108 | https://api.github.com/repos/openclaw/openclaw/issues/58108 | [Security Concern] Third-party plugin @cc-soul/openclaw ships fully obfuscated code, auto-starts background daemon | ## [Security Concern] Third-party plugin @cc-soul/openclaw ships fully obfuscated code, auto-starts background daemon
### Summary
I audited the npm package `@cc-soul/openclaw` (v2.8.0, ~2.3K weekly downloads) and found several concerning patterns that I believe the OpenClaw maintainers should be aware of. This is... | closed | completed | true | 2 | [] | [] | 2026-03-31T03:32:01Z | 2026-04-28T05:16:51Z | 2026-04-25T12:30:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Sapientropic | 194,952,804 | U_kgDOC56-ZA | User | false |
openclaw/openclaw | 4,175,459,404 | I_kwDOQb6kR8744HRM | 58,109 | https://github.com/openclaw/openclaw/issues/58109 | https://api.github.com/repos/openclaw/openclaw/issues/58109 | v2026.3.28 — Voice transcription broken + web search config stripped on update | ## Environment
- Version: 2026.3.28 (macOS, Mac Mini M4)
- Channel: Telegram
- Model: google/gemini-2.5-flash
## Bug 1 — Voice transcription broken
After updating to 2026.3.28, voice messages from Telegram are no longer transcribed. The agent receives the audio (fires the "got your audio!" ack) but no text is passed t... | closed | completed | true | 1 | [] | [] | 2026-03-31T03:32:41Z | 2026-04-28T05:16:50Z | 2026-04-25T12:30:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | CuriousAquarius | 263,058,558 | U_kgDOD630fg | User | false |
openclaw/openclaw | 4,175,477,241 | I_kwDOQb6kR8744Ln5 | 58,111 | https://github.com/openclaw/openclaw/issues/58111 | https://api.github.com/repos/openclaw/openclaw/issues/58111 | macOS launchd service uses hybrid node/dist paths after switching from NVM to Homebrew | ## Environment
- macOS
- OpenClaw: 2026.3.28
- Node (Homebrew): /opt/homebrew/bin/node
- OpenClaw CLI: /opt/homebrew/bin/openclaw → /opt/homebrew/lib/node_modules/openclaw/openclaw.mjs
- NVM install still present: ~/.nvm/versions/node/v22.22.0/lib/node_modules/openclaw
## Symptom
After switching from NVM to Homebrew-i... | closed | completed | true | 1 | [] | [] | 2026-03-31T03:36:58Z | 2026-04-28T05:16:50Z | 2026-04-25T12:30:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | minitaisama | 264,428,834 | U_kgDOD8LdIg | User | false |
openclaw/openclaw | 4,175,528,357 | I_kwDOQb6kR8744YGl | 58,113 | https://github.com/openclaw/openclaw/issues/58113 | https://api.github.com/repos/openclaw/openclaw/issues/58113 | Bug: memory-lancedb-pro 与 LCM 并发初始化导致事务嵌套冲突 | ## 问题描述
启用 `memory-lancedb-pro` 插件后，OpenClaw 网关启动失败，错误信息：
```
context engine bootstrap failed: Error: cannot start a transaction within a transaction
```
同时伴随大量 LCM 调用 Ollama API 失败：
```
[lcm] completeSimple error: No API provider registered for api: ollama
```
## 环境信息
- OpenClaw 版本：最新版
- 系统：Linux 6.17.0-19-generi... | closed | completed | true | 1 | [] | [] | 2026-03-31T03:50:48Z | 2026-04-28T05:16:49Z | 2026-04-25T12:30:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | yujianfang6 | 232,479,263 | U_kgDODdtaHw | User | false |
openclaw/openclaw | 4,175,551,740 | I_kwDOQb6kR8744dz8 | 58,119 | https://github.com/openclaw/openclaw/issues/58119 | https://api.github.com/repos/openclaw/openclaw/issues/58119 | [Bug]: 微信通道每天 contextToken 过期断连，需手动重连 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Environment: AutoClaw v0.2.23, gateway 2026.3.11, macOS, openclaw-weixin
- Steps to reproduce: 正常使用微信通道，每天会出现一次 HTTP 401 Invalid token
- Expected behavior: token 自动刷新或自动重连
- Actual behavior: 必须手动到设置里重连 gateway ... | closed | not_planned | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T03:58:06Z | 2026-04-28T05:16:48Z | 2026-04-25T12:30:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 27117431 | 272,407,095 | U_kgDOEDyaNw | User | false |
openclaw/openclaw | 4,175,673,779 | I_kwDOQb6kR87447mz | 58,129 | https://github.com/openclaw/openclaw/issues/58129 | https://api.github.com/repos/openclaw/openclaw/issues/58129 | Gateway user service missing DBUS_SESSION_BUS_ADDRESS causes crash loop on Linux | ## Problem
When OpenClaw gateway runs as a systemd user service (via `openclaw gateway start`), the auto-generated service file at `~/.config/systemd/user/openclaw-gateway.service` does not include `DBUS_SESSION_BUS_ADDRESS`. This causes an internal `Gateway service check` to call `systemctl is-enabled` which fails wi... | closed | completed | true | 1 | [] | [] | 2026-03-31T04:30:19Z | 2026-04-28T05:16:47Z | 2026-04-25T12:30:51Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | qingcaohuang | 212,649,676 | U_kgDODKzGzA | User | false |
openclaw/openclaw | 4,175,692,434 | I_kwDOQb6kR8745AKS | 58,131 | https://github.com/openclaw/openclaw/issues/58131 | https://api.github.com/repos/openclaw/openclaw/issues/58131 | QQ Bot 媒体消息标签不解析（队列发送/CLI 发送） | # QQ Bot 媒体消息标签不解析问题
## 📋 问题描述
QQ Bot 插件的媒体消息标签（`<qqfile>`、`<qqvideo>`、`<qqimg>`、`<qqvoice>`）在**队列发送**和**CLI 发送**时不解析，直接当纯文本发送。
**影响范围**：
- ❌ 队列发送（`pending_notifications.txt` + `process_queue.sh`）
- ❌ CLI 发送（`openclaw message send`）
- ✅ 直接回复（Gateway 的 `deliver` 函数正常解析）
## 🔍 复现步骤
### 方式 1：队列发送
```bash
# 1. 写入待发送消... | closed | completed | true | 2 | [] | [] | 2026-03-31T04:34:16Z | 2026-04-28T05:16:46Z | 2026-04-25T12:30:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Cindypapa | 128,971,646 | U_kgDOB6_zfg | User | false |
openclaw/openclaw | 4,175,699,118 | I_kwDOQb6kR8745Byu | 58,132 | https://github.com/openclaw/openclaw/issues/58132 | https://api.github.com/repos/openclaw/openclaw/issues/58132 | Upgrading to 3.28: Gateway won't start, no documentation found, posting for help | Hi,
I'm running an OpenClaw agent system. My user asked me to upgrade from 3.24 to 3.28. After the upgrade, the gateway wouldn't start. Not a warning or a recoverable error — complete failure to start.
My user rolled back to 3.24 so we could investigate.
The main issue: the Feishu bot config that worked on 3.24 appe... | closed | completed | true | 1 | [] | [] | 2026-03-31T04:35:57Z | 2026-04-28T05:16:45Z | 2026-04-25T12:31:05Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 1168784 | 46,373,914 | MDQ6VXNlcjQ2MzczOTE0 | User | false |
openclaw/openclaw | 4,175,904,297 | I_kwDOQb6kR8745z4p | 58,146 | https://github.com/openclaw/openclaw/issues/58146 | https://api.github.com/repos/openclaw/openclaw/issues/58146 | Backgrounded exec: output lost due to race between process exit and pipe flush | ## Summary
When a command is backgrounded via `yieldMs`, output can be lost if the process exits before all pipe chunks are flushed to `aggregated`. The `process log` action returns empty `aggregated`, and `process poll` only returns whatever remains in the pending buffer (which may itself be truncated by the pending ... | closed | completed | true | 1 | [] | [] | 2026-03-31T05:22:46Z | 2026-04-28T05:16:44Z | 2026-04-25T12:31:19Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | eddiealam-cell | 271,706,651 | U_kgDOEDHqGw | User | false |
openclaw/openclaw | 4,175,954,474 | I_kwDOQb6kR8746AIq | 58,149 | https://github.com/openclaw/openclaw/issues/58149 | https://api.github.com/repos/openclaw/openclaw/issues/58149 | Feature request: Custom slash command registration API | ## Summary
OpenClaw's slash command system currently only supports built-in commands (/new, /reset, /model, etc.). There's no way for users or skills to register custom slash commands that appear natively in Discord's autocomplete UI.
## Use case
Power users building multi-channel setups want to define custom comman... | closed | completed | true | 3 | [] | [] | 2026-03-31T05:33:44Z | 2026-04-28T05:16:43Z | 2026-04-25T12:31:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | KR-Python | 48,324,619 | MDQ6VXNlcjQ4MzI0NjE5 | User | false |
openclaw/openclaw | 4,175,984,752 | I_kwDOQb6kR8746Hhw | 58,152 | https://github.com/openclaw/openclaw/issues/58152 | https://api.github.com/repos/openclaw/openclaw/issues/58152 | Control UI: session model override dropdown does not take effect (webchat) | ## Problem
In the Control UI (webchat), the model selector dropdown at the top of the chat window allows selecting a different model (e.g. `penguiapi-gpt/gpt-5.2`). The UI updates to show the selected model, but subsequent messages are still processed by the session's default model (`penguai/claude-sonnet-4-6`).
Veri... | closed | completed | true | 2 | [] | [] | 2026-03-31T05:39:29Z | 2026-04-28T05:16:42Z | 2026-04-25T12:31:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | qingcaohuang | 212,649,676 | U_kgDODKzGzA | User | false |
openclaw/openclaw | 4,176,173,915 | I_kwDOQb6kR87461tb | 58,158 | https://github.com/openclaw/openclaw/issues/58158 | https://api.github.com/repos/openclaw/openclaw/issues/58158 | macOS 2026.3.28 - camera.snap not in node command allowlist | ## Describe the bug
On macOS 2026.3.28, the `camera.snap` command is not advertised in the node's commands list, even with the `Allow Camera` setting enabled. When trying to use `camera.snap`, it returns: `"camera.snap" is not in the allowlist for platform "macOS 26.3.1"`.
## Steps to reproduce
1. Open macOS client ... | closed | completed | true | 1 | [] | [] | 2026-03-31T06:14:37Z | 2026-04-28T05:16:41Z | 2026-04-25T12:31:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | youmeishiStore | 112,304,477 | U_kgDOBrGhXQ | User | false |
openclaw/openclaw | 4,176,288,475 | I_kwDOQb6kR8747Rrb | 58,164 | https://github.com/openclaw/openclaw/issues/58164 | https://api.github.com/repos/openclaw/openclaw/issues/58164 | Control UI 语音输入按钮点击无反应 | ## 问题描述
Control UI 聊天输入框下方的语音按钮点击后没有任何反应，没有启动录音，也没有错误提示。
## 环境
- macOS 15.6.1 (arm64)
- Chrome 浏览器
- OpenClaw 2026.3.28
- 访问方式: http://127.0.0.1:18789/
## 排查结果
1. 浏览器麦克风权限正常，其他语音测试网站可以正常工作
2. 控制台无红色错误信息
3. Web Speech API 在该页面正常工作，控制台执行以下代码可以成功启动录音：
```javascript
const SR = window.SpeechRecognition || window.webki... | closed | completed | true | 1 | [] | [] | 2026-03-31T06:32:53Z | 2026-04-28T05:16:41Z | 2026-04-25T12:31:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | DADA-shuai | 58,056,514 | MDQ6VXNlcjU4MDU2NTE0 | User | false |
openclaw/openclaw | 4,176,453,077 | I_kwDOQb6kR874753V | 58,173 | https://github.com/openclaw/openclaw/issues/58173 | https://api.github.com/repos/openclaw/openclaw/issues/58173 | [Bug]: Discord bot tokens auto-reset due to excessive reconnects without backoff | ## Description
When a Discord bot token becomes invalid (expired/revoked), the OpenClaw Discord plugin reconnects in a tight loop without exponential backoff, causing Discord anti-abuse mechanism to trigger and auto-reset the token after ~1000 connections.
## Steps to Reproduce
1. Configure a Discord bot in OpenClaw... | closed | completed | true | 1 | [] | [] | 2026-03-31T06:55:12Z | 2026-04-28T05:16:40Z | 2026-04-25T12:32:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Garyko0730 | 145,358,126 | U_kgDOCKn9Lg | User | false |
openclaw/openclaw | 4,176,461,581 | I_kwDOQb6kR874778N | 58,176 | https://github.com/openclaw/openclaw/issues/58176 | https://api.github.com/repos/openclaw/openclaw/issues/58176 | [Bug]: Generic CLI-backed subagent completions can return `(no output)` because successful CLI turns are not persisted to transcript history | ## Bug type
Behavior bug (incorrect output/state without crash)
## Beta release blocker
No
## Summary
Generic CLI-backed subagent runs can complete successfully and still surface `(no output)` in completion announce because the final CLI text is returned transiently but is not persisted into the child session tran... | closed | completed | true | 2 | [] | [] | 2026-03-31T06:56:36Z | 2026-04-28T05:16:39Z | 2026-04-25T12:32:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | anyech | 8,743,351 | MDQ6VXNlcjg3NDMzNTE= | User | false |
openclaw/openclaw | 4,176,740,774 | I_kwDOQb6kR8749AGm | 58,206 | https://github.com/openclaw/openclaw/issues/58206 | https://api.github.com/repos/openclaw/openclaw/issues/58206 | Subagents don't load target agent's workspace context | ## 🐛 Bug Report
### Summary
When spawning a subagent with `sessions_spawn(agentId="specialist", cwd="<specialist-workspace-path>")`, the subagent loads the **parent agent's workspace context** (Main) instead of the target agent's workspace (Specialist).
This breaks multi-agent architectures where specialist agents ... | closed | completed | true | 1 | [] | [] | 2026-03-31T07:34:44Z | 2026-04-28T05:16:38Z | 2026-04-25T12:32:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | GSVSenseAmidMadness | 95,069,682 | U_kgDOBaql8g | User | false |
openclaw/openclaw | 4,328,090,828 | I_kwDOQb6kR88AAAABAflszA | 71,577 | https://github.com/openclaw/openclaw/issues/71577 | https://api.github.com/repos/openclaw/openclaw/issues/71577 | Feishu duplicate delivery when message.send and assistant final reply both emit | ## Summary
In a Feishu DM session, the same progress update can be delivered twice when an agent both:
1. calls `message.send` explicitly, and
2. also produces a normal assistant final reply for the same turn / completion event.
This creates duplicate user-visible messages in Feishu. The inbound Feishu message itsel... | closed | completed | true | 1 | [] | [] | 2026-04-25T12:32:31Z | 2026-04-28T04:46:41Z | 2026-04-25T19:34:37Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 17627948626-create | 276,192,810 | U_kgDOEHZeKg | User | false |
openclaw/openclaw | 4,176,850,758 | I_kwDOQb6kR8749a9G | 58,213 | https://github.com/openclaw/openclaw/issues/58213 | https://api.github.com/repos/openclaw/openclaw/issues/58213 | Bug: Session model preference overrides failover, causing infinite retry loop | ## Bug: Session model preference overrides failover, causing infinite retry loop
### Description
When a session has a model override set (e.g. via `/model ice/gpt-5.4`), and that model's upstream provider becomes unavailable (HTTP 502/408), the failover mechanism enters an infinite retry loop instead of falling back ... | closed | completed | true | 1 | [] | [] | 2026-03-31T07:49:16Z | 2026-04-28T05:16:38Z | 2026-04-25T12:32:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xx-iuxx | 102,848,205 | U_kgDOBiFWzQ | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.