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 | 4,205,170,602 | I_kwDOQb6kR876pc-q | 60,921 | https://github.com/openclaw/openclaw/issues/60921 | https://api.github.com/repos/openclaw/openclaw/issues/60921 | [Bug] Isolated cron session delivery fails with 'Channel is required' even when delivery.channel is explicitly set | ## Bug Description
When a cron job uses `sessionTarget: "isolated"` with `delivery.mode: "announce"` and explicitly sets `delivery.channel`, `delivery.to`, `delivery.accountId`, and `delivery.bestEffort: true`, the delivery still fails with:
```
Channel is required (no configured channels detected). Set delivery.chan... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T15:25:49Z | 2026-04-28T05:10:10Z | 2026-04-25T13:48:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | qq475059648 | 42,694,909 | MDQ6VXNlcjQyNjk0OTA5 | User | false |
openclaw/openclaw | 4,205,203,288 | I_kwDOQb6kR876pk9Y | 60,926 | https://github.com/openclaw/openclaw/issues/60926 | https://api.github.com/repos/openclaw/openclaw/issues/60926 | [Bug]: Heartbeat injects into active sub-agent sessions, terminating them with HEARTBEAT_OK | ## Description
Heartbeat prompts are being injected into **active sub-agent sessions** (`agent:<id>:subagent:<uuid>`), causing the sub-agent to respond `HEARTBEAT_OK` instead of completing its actual task. This effectively kills any sub-agent run that spans a heartbeat interval.
## Steps to Reproduce
1. Configure he... | closed | completed | true | 2 | [] | [] | 2026-04-04T15:41:55Z | 2026-04-28T05:10:09Z | 2026-04-25T13:48:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | drmagnuson | 180,776,376 | U_kgDOCsZtuA | User | false |
openclaw/openclaw | 4,205,210,776 | I_kwDOQb6kR876pmyY | 60,929 | https://github.com/openclaw/openclaw/issues/60929 | https://api.github.com/repos/openclaw/openclaw/issues/60929 | sessions_spawn subagent fails with 1008 pairing required - never enters pending queue | null | closed | not_planned | true | 3 | [] | [] | 2026-04-04T15:45:43Z | 2026-04-28T05:10:09Z | 2026-04-25T13:48:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | meqte | 105,053,344 | U_kgDOBkL8oA | User | false |
openclaw/openclaw | 4,205,227,489 | I_kwDOQb6kR876pq3h | 60,933 | https://github.com/openclaw/openclaw/issues/60933 | https://api.github.com/repos/openclaw/openclaw/issues/60933 | [Proposal] Prevent order-sensitive tool failures with a pre-execution logic fuse | ### Summary
## Problem
Some agent failures are not permission failures — they are **execution-order failures**.
Example:
- `rm -rf /data`
- `backup /data/logs`
Both actions may be individually allowed.
But if they run in the wrong order, the backup fails and state is lost.
This is a gap that RBAC, locks, and st... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-04T15:54:23Z | 2026-04-28T05:10:08Z | 2026-04-25T13:48:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ZC502 | 218,752,469 | U_kgDODQnl1Q | User | false |
openclaw/openclaw | 4,205,273,973 | I_kwDOQb6kR876p2N1 | 60,940 | https://github.com/openclaw/openclaw/issues/60940 | https://api.github.com/repos/openclaw/openclaw/issues/60940 | iMessage (imsg): sent messages with is_from_me=true echoed back as inbound to session | ## Bug Report
**Version:** OpenClaw 2026.4.1 (da64a97), imsg 0.5.0
**Channel:** iMessage (legacy imsg integration)
## Description
Outbound messages sent by the gateway (my replies) are being fed back into the session as inbound messages, causing an infinite echo loop.
## Root Cause
The `imsg` JSON-RPC watch strea... | closed | completed | true | 1 | [] | [] | 2026-04-04T16:17:20Z | 2026-04-28T05:10:06Z | 2026-04-25T13:49:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | deliogdos7 | 271,852,146 | U_kgDOEDQicg | User | false |
openclaw/openclaw | 4,205,277,562 | I_kwDOQb6kR876p3F6 | 60,941 | https://github.com/openclaw/openclaw/issues/60941 | https://api.github.com/repos/openclaw/openclaw/issues/60941 | Subagent/cron completion events route to wrong sender's session (cross-contamination) | ## Bug Description
When a subagent or cron task completes while the agent is mid-turn handling a different user's request, the completion output routes to the **currently active requester** instead of the **originating session/sender**.
This causes two user-visible symptoms:
1. **NO_REPLY token leaking** — the agent... | closed | completed | true | 1 | [] | [] | 2026-04-04T16:19:01Z | 2026-04-28T05:10:06Z | 2026-04-25T13:49:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | swryder | 1,392,120 | MDQ6VXNlcjEzOTIxMjA= | User | false |
openclaw/openclaw | 4,205,283,860 | I_kwDOQb6kR876p4oU | 60,943 | https://github.com/openclaw/openclaw/issues/60943 | https://api.github.com/repos/openclaw/openclaw/issues/60943 | Feature: Pre/Post Tool Use Hooks (programmable safety rails) | ## Summary
Add configurable shell-script hooks that run before and after every tool execution, allowing users to programmatically approve, deny, or augment tool calls.
## Motivation
Claude Code's architecture (and the Claw Code clean-room rewrite) implements a hook system where shell commands run at PreToolUse and Pos... | closed | not_planned | true | 5 | [] | [] | 2026-04-04T16:22:19Z | 2026-04-28T05:10:05Z | 2026-04-25T13:49:16Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ohmanthatsgreat | 269,280,971 | U_kgDOEAzmyw | User | false |
openclaw/openclaw | 4,205,284,480 | I_kwDOQb6kR876p4yA | 60,944 | https://github.com/openclaw/openclaw/issues/60944 | https://api.github.com/repos/openclaw/openclaw/issues/60944 | Feature: Proactive auto-compaction with chained summaries | ## Summary
Automatically compact session history when approaching context limits, using a summary-chaining strategy that preserves deep context across multiple compactions.
## Motivation
Currently, sessions hit context limits unexpectedly, requiring manual `/compact`. Claude Code's architecture reveals a robust multi-... | closed | completed | true | 1 | [] | [] | 2026-04-04T16:22:37Z | 2026-04-28T05:10:04Z | 2026-04-25T13:49:23Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ohmanthatsgreat | 269,280,971 | U_kgDOEAzmyw | User | false |
openclaw/openclaw | 4,205,284,958 | I_kwDOQb6kR876p45e | 60,945 | https://github.com/openclaw/openclaw/issues/60945 | https://api.github.com/repos/openclaw/openclaw/issues/60945 | Feature: Subagent tool allowlists (per-agent-type tool sandboxing) | ## Summary
Allow configuring which tools each sub-agent type has access to, preventing spawned agents from having broader capabilities than needed.
## Motivation
Currently, sub-agents spawned via `sessions_spawn` inherit tool access from their parent or get the full tool set. Claude Code and Claw Code both implement e... | closed | completed | true | 1 | [] | [] | 2026-04-04T16:22:53Z | 2026-04-28T05:10:03Z | 2026-04-25T13:49:30Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ohmanthatsgreat | 269,280,971 | U_kgDOEAzmyw | User | false |
openclaw/openclaw | 4,205,291,250 | I_kwDOQb6kR876p6by | 60,947 | https://github.com/openclaw/openclaw/issues/60947 | https://api.github.com/repos/openclaw/openclaw/issues/60947 | Feature: Degraded mode for model fallback — prevent death spiral when cloud model is throttled | ## Summary
When the primary cloud model is throttled or unavailable and OpenClaw falls back to a smaller local model, the fallback model receives the **same full context** (often 100K+ tokens) that was designed for a large cloud model. This creates a death spiral:
1. Cloud model returns 429/overloaded → fallback to s... | closed | completed | true | 1 | [] | [] | 2026-04-04T16:26:29Z | 2026-04-28T05:10:03Z | 2026-04-25T13:49:37Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | andyk-ms | 91,510,251 | U_kgDOBXRV6w | User | false |
openclaw/openclaw | 4,205,291,351 | I_kwDOQb6kR876p6dX | 60,948 | https://github.com/openclaw/openclaw/issues/60948 | https://api.github.com/repos/openclaw/openclaw/issues/60948 | [BUG] Agent Session 超时不自动恢复，导致 sessions_send 持续 timeout | ## 问题描述
Agent Session 空闲超时后会 ended，之后对该 Session 发送 sessions_send 会返回 timeout 错误，需要手动重启 Gateway 才能恢复。
## 复现步骤
1. Agent Session 运行一段时间后空闲超时
2. Session 状态变为 ended
3. 调用 sessions_send(sessionId, message) → timeout
4. 无法自动恢复，只能手动 restart Gateway
## 期望行为
- Session 超时后应该能自动重新激活（类似人重新发消息触发）
- 或者有 Keep-alive 机制保活
- 或者 sess... | closed | completed | true | 1 | [] | [] | 2026-04-04T16:26:33Z | 2026-04-28T05:10:02Z | 2026-04-25T13:49:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ROLLcatCLUB | 151,552,329 | U_kgDOCQiBSQ | User | false |
openclaw/openclaw | 4,205,405,148 | I_kwDOQb6kR876qWPc | 60,973 | https://github.com/openclaw/openclaw/issues/60973 | https://api.github.com/repos/openclaw/openclaw/issues/60973 | neodata skill query.py proxy request returns 401 — auth token not injected | **Environment:**
- OpenClaw `2026.3.12`
- neodata 金融数据技能 (`query.py`)
- Client session authenticated
**Issue:**
`query.py` calls `http://localhost:19000/proxy/api` and consistently returns `401 Unauthorized`, even though the client is logged in.
**Suspected Root Cause:**
Auth Gateway does not correctly inject the ses... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T17:23:43Z | 2026-04-28T05:10:00Z | 2026-04-25T13:50:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hzq1957946 | 248,799,445 | U_kgDODtRg1Q | User | false |
openclaw/openclaw | 4,205,410,149 | I_kwDOQb6kR876qXdl | 60,975 | https://github.com/openclaw/openclaw/issues/60975 | https://api.github.com/repos/openclaw/openclaw/issues/60975 | [Feature]: Built-in deterministic cron receipt ledger (machine-trusted JSONL) + optional markdown human receipt split | ### Summary
When using cron-driven memory promotion/self-learning workflows, model-written receipts are semantically useful but not strict enough for machine validation. We need a first-class deterministic ledger emitted by OpenClaw control-plane/runtime, separate from model text.
### Problem to solve
Today, if user... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-04T17:26:23Z | 2026-04-28T05:10:00Z | 2026-04-25T13:50:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | aemeathpa | 269,386,194 | U_kgDOEA6B0g | User | false |
openclaw/openclaw | 4,205,414,868 | I_kwDOQb6kR876qYnU | 60,976 | https://github.com/openclaw/openclaw/issues/60976 | https://api.github.com/repos/openclaw/openclaw/issues/60976 | Feature Request: Native Multi-Agent Team Mode (Like Claude Code Agent Teams) | ## Current Pain Points
### 1. No Native Agent-to-Agent Collaboration
The current architecture uses a star topology where all sub-agents communicate only through the gateway with the main agent. Sub-agents cannot:
- Talk directly to each other
- Have discussions, debates, or brainstorming sessions
- Autonomously divid... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T17:29:00Z | 2026-04-28T05:09:59Z | 2026-04-25T13:50:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | davidme6 | 198,009,354 | U_kgDOC81iCg | User | false |
openclaw/openclaw | 4,328,234,409 | I_kwDOQb6kR88AAAABAfudqQ | 71,603 | https://github.com/openclaw/openclaw/issues/71603 | https://api.github.com/repos/openclaw/openclaw/issues/71603 | Cron may keep stale in-memory schedule after job expression change, causing duplicate runs | ## Summary
After migrating OpenClaw cron config from a single daily Instagram editorial job to split weekday/weekend jobs, the old daily schedule appears to have remained active in memory. On Saturday, April 25, 2026, the old daily job fired at 06:00 America/Sao_Paulo and the new weekend job fired at 07:00, causing two... | closed | completed | true | 2 | [] | [] | 2026-04-25T13:40:23Z | 2026-04-28T05:09:58Z | 2026-04-25T13:50:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rafaelfaleiro | 6,382,616 | MDQ6VXNlcjYzODI2MTY= | User | false |
openclaw/openclaw | 4,205,432,401 | I_kwDOQb6kR876qc5R | 60,980 | https://github.com/openclaw/openclaw/issues/60980 | https://api.github.com/repos/openclaw/openclaw/issues/60980 | [Bug]: Channels table empty and Telegram/Discord unavailable on Windows despite valid config and installed deps | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
On Windows with OpenClaw 2026.4.2, the gateway is reachable but the active channel registry remains empty, so outbound Telegram and Discord sends fail with Channel is unavailable despite valid config and installed runtime de... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-04T17:37:00Z | 2026-04-28T05:09:57Z | 2026-04-25T13:50:47Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kirilli40 | 176,934,438 | U_kgDOCovOJg | User | false |
openclaw/openclaw | 4,205,454,648 | I_kwDOQb6kR876qiU4 | 60,983 | https://github.com/openclaw/openclaw/issues/60983 | https://api.github.com/repos/openclaw/openclaw/issues/60983 | Cron announce delivery fails with 'message is too long' on Telegram — no auto-chunking | When a cron job with `delivery: { mode: "announce" }` produces output exceeding 4,096 characters, the Telegram delivery fails with:
```
GrammyError: Call to 'sendMessage' failed! (400: Bad Request: message is too long)
```
The normal outbound reply path already chunks via `splitTelegramHtmlChunks` / `sendTelegramText... | closed | completed | true | 1 | [] | [] | 2026-04-04T17:47:35Z | 2026-04-28T05:09:57Z | 2026-04-25T13:50:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Warningsigns1 | 136,926,920 | U_kgDOCClWyA | User | false |
openclaw/openclaw | 4,205,467,781 | I_kwDOQb6kR876qliF | 60,988 | https://github.com/openclaw/openclaw/issues/60988 | https://api.github.com/repos/openclaw/openclaw/issues/60988 | MiniMax Portal OAuth: API requests return nginx 404 even with valid token and model config | ## Bug
### Summary
Even after fixing the empty `models: []` config issue (#60987), requests to `minimax-portal/MiniMax-M2.7-highspeed` through the gateway return a bare nginx `404 Not Found` and fall back to the next provider. The same OAuth token works perfectly when calling the MiniMax API directly via curl.
### Ev... | closed | completed | true | 2 | [] | [] | 2026-04-04T17:53:33Z | 2026-04-28T05:09:56Z | 2026-04-25T13:51:01Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chenxin-yan | 71,162,231 | MDQ6VXNlcjcxMTYyMjMx | User | false |
openclaw/openclaw | 4,205,473,525 | I_kwDOQb6kR876qm71 | 60,989 | https://github.com/openclaw/openclaw/issues/60989 | https://api.github.com/repos/openclaw/openclaw/issues/60989 | runtime.modelAuth doesn't handle GitHub Copilot token exchange — breaks all plugins using Copilot | ## Summary
`runtime.modelAuth.getApiKeyForModel()` returns the raw GitHub user token (`ghu_...`) for the `github-copilot` provider, but the Copilot API requires an exchanged API token (`tid=...`) obtained via `https://api.github.com/copilot_internal/v2/token`.
OpenClaw core handles this exchange internally in `agent-... | closed | completed | true | 1 | [] | [] | 2026-04-04T17:55:56Z | 2026-04-28T05:09:55Z | 2026-04-25T13:51:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bradfreels | 179,389,909 | U_kgDOCrFF1Q | User | false |
openclaw/openclaw | 4,205,491,154 | I_kwDOQb6kR876qrPS | 60,992 | https://github.com/openclaw/openclaw/issues/60992 | https://api.github.com/repos/openclaw/openclaw/issues/60992 | [Feature]: web_fetch should allow private IP access for users behind TUN/proxy (Fake-IP DNS issue) | ## Feature Request
### Problem
When Chinese users (and many others) use a **TUN mode proxy** (V2RayN, Clash, etc.), the DNS resolver returns **Fake-IP** addresses (e.g., `198.18.x.x`, `127.0.0.1`) for all domains. This is a standard transparent proxy behavior.
When using `web_fetch` to access any URL in this scenari... | closed | completed | true | 1 | [] | [] | 2026-04-04T18:03:31Z | 2026-04-28T05:09:54Z | 2026-04-25T13:51:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | prospect1314521 | 135,647,789 | U_kgDOCBXSLQ | User | false |
openclaw/openclaw | 4,205,585,782 | I_kwDOQb6kR876rCV2 | 61,007 | https://github.com/openclaw/openclaw/issues/61007 | https://api.github.com/repos/openclaw/openclaw/issues/61007 | [Bug] Gemini models displayed as anthropic in openclaw models list | ## Description
When running `openclaw models list`, Gemini models are incorrectly displayed with the `anthropic` provider prefix instead of `google`.
## Steps to Reproduce
1. Configure Google/Gemini models in openclaw.json under `models.providers.google`
2. Run `openclaw models list`
3. Observe that Gemini models show... | closed | completed | true | 1 | [] | [] | 2026-04-04T18:50:50Z | 2026-04-28T05:09:54Z | 2026-04-25T13:51:29Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rsugumar | 1,397,811 | MDQ6VXNlcjEzOTc4MTE= | User | false |
openclaw/openclaw | 4,205,677,055 | I_kwDOQb6kR876rYn_ | 61,015 | https://github.com/openclaw/openclaw/issues/61015 | https://api.github.com/repos/openclaw/openclaw/issues/61015 | [Bug] cron rm command hangs indefinitely with no output or error | ## Bug Description
`openclaw cron rm <id>` hangs indefinitely with no output, even with valid job IDs. The command never returns and must be killed manually with Ctrl+C.
## Steps to Reproduce
1. Run `openclaw cron rm <valid-job-id>`
2. Command hangs with no output, no error, no prompt
3. Must use `Ctrl+C` to recover
... | closed | completed | true | 1 | [] | [] | 2026-04-04T19:41:31Z | 2026-04-28T05:09:53Z | 2026-04-25T13:51:36Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | karlhillx | 13,650,232 | MDQ6VXNlcjEzNjUwMjMy | User | false |
openclaw/openclaw | 4,205,703,699 | I_kwDOQb6kR876rfIT | 61,018 | https://github.com/openclaw/openclaw/issues/61018 | https://api.github.com/repos/openclaw/openclaw/issues/61018 | Feature request: cross-channel relay/notification for inbound DMs in per-channel-peer mode | ## Problem
When `session.dmScope` is set to `per-channel-peer`, inbound DMs on non-primary channels (e.g., WhatsApp) create isolated sessions. The agent in the main session (e.g., Slack) has no visibility into these inbound messages.
## Real-world scenario
1. User instructs agent (via Slack) to send a WhatsApp messa... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T19:57:25Z | 2026-04-28T05:09:52Z | 2026-04-25T13:51:43Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | redasadki | 5,021,946 | MDQ6VXNlcjUwMjE5NDY= | User | false |
openclaw/openclaw | 4,205,737,497 | I_kwDOQb6kR876rnYZ | 61,025 | https://github.com/openclaw/openclaw/issues/61025 | https://api.github.com/repos/openclaw/openclaw/issues/61025 | UI Bug: Assistant avatar rendered inline instead of as message bubble icon | ## Description
The assistant avatar (configured in `IDENTITY.md` and present in the filesystem) is incorrectly rendered in the Control UI. Instead of appearing as an icon next to the assistant's message bubble (in the message stream), it is being rendered as an inline element inside the text response itself, typically ... | closed | completed | true | 1 | [] | [] | 2026-04-04T20:18:15Z | 2026-04-28T05:09:51Z | 2026-04-25T13:51:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ekamsat | 12,574,123 | MDQ6VXNlcjEyNTc0MTIz | User | false |
openclaw/openclaw | 4,205,756,830 | I_kwDOQb6kR876rsGe | 61,027 | https://github.com/openclaw/openclaw/issues/61027 | https://api.github.com/repos/openclaw/openclaw/issues/61027 | llm-task: thinking param not auto-applied for reasoning models (Gemini 2.5 Pro fails with Budget 0) | ## Bug
When calling `llm-task` with a reasoning model like `google/gemini-2.5-pro` (which has `reasoning: true` in the model config), and the caller does **not** explicitly pass the `thinking` parameter, the embedded runner sends `thinkingBudget: 0` to the Google Generative AI API.
Gemini 2.5 Pro rejects this with:
`... | closed | completed | true | 1 | [] | [] | 2026-04-04T20:31:23Z | 2026-04-28T05:09:50Z | 2026-04-25T13:52:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Maz0327 | 99,418,454 | U_kgDOBe0BVg | User | false |
openclaw/openclaw | 4,205,787,973 | I_kwDOQb6kR876rztF | 61,031 | https://github.com/openclaw/openclaw/issues/61031 | https://api.github.com/repos/openclaw/openclaw/issues/61031 | 微信渠道 TTS 语音消息应发送为原生语音格式（silk）而非文件 | ## 问题描述
当 OpenClaw 通过 TTS 生成语音并发送到微信渠道时，语音以 MP3 文件形式发送（文件卡片），而不是微信原生语音消息（圆形气泡）。
## 期望行为
TTS 音频应自动转码为 SILK 格式，通过微信原生语音接口发送，显示为微信语音气泡。
## 现状分析
- 微信 plugin (`openclaw-weixin`) 已有 `silk-transcode.js`，但只实现了 SILK → WAV 解码（用于接收语音），没有 WAV → SILK 编码（用于发送）
- `silk-wasm` npm 包同时支持 encode/decode，可作为编码方案
- 微信 plugin 的 `sendWei... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T20:51:40Z | 2026-04-28T05:09:50Z | 2026-04-25T13:52:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hchwork | 2,044,007 | MDQ6VXNlcjIwNDQwMDc= | User | false |
openclaw/openclaw | 4,205,815,965 | I_kwDOQb6kR876r6id | 61,037 | https://github.com/openclaw/openclaw/issues/61037 | https://api.github.com/repos/openclaw/openclaw/issues/61037 | [Bug]: Matrix infinite self-message loop when verboseDefault is set to full | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Setting `agents.defaults.verboseDefault` to `"full"` causes the Matrix channel to enter an infinite message loop: tool output sent as `m.text` arrives back via `/sync`, passes through the event listener uncondi... | closed | completed | true | 1 | [] | [] | 2026-04-04T21:09:44Z | 2026-04-28T05:09:49Z | 2026-04-25T13:52:25Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | esrehmki | 20,036,971 | MDQ6VXNlcjIwMDM2OTcx | User | false |
openclaw/openclaw | 4,205,825,342 | I_kwDOQb6kR876r80- | 61,039 | https://github.com/openclaw/openclaw/issues/61039 | https://api.github.com/repos/openclaw/openclaw/issues/61039 | Feature request: per-session workspace for multi-agent channel isolation | ## 需求
当微信 plugin 连接多个微信机器人时，每个机器人的会话应该有独立的工作空间（workspace），文件和记忆互不干扰。
## 当前行为
所有微信会话共享一个全局 workspace（`~/.openclaw/workspace`），所有 session 的文件、记忆、skills 都在一起，无法隔离。
## 期望行为
每个微信会话/机器人自动创建独立 workspace，实现：
- 独立的文件存储
- 独立的记忆（MEMORY.md）
- 独立的 session 历史
## 适用场景
- 一个 OpenClaw 实例管理多个客户的微信机器人
- 每个机器人的上下文、文件、记忆需要完全隔离
- 避免不同... | closed | completed | true | 1 | [] | [] | 2026-04-04T21:15:47Z | 2026-04-28T05:09:48Z | 2026-04-25T13:52:32Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hchwork | 2,044,007 | MDQ6VXNlcjIwNDQwMDc= | User | false |
openclaw/openclaw | 4,205,837,121 | I_kwDOQb6kR876r_tB | 61,042 | https://github.com/openclaw/openclaw/issues/61042 | https://api.github.com/repos/openclaw/openclaw/issues/61042 | Plugin code-safety scanner falsely blocks openclaw-quiubo update on bundled child_process / HOME path patterns | ## Summary
`openclaw plugins update openclaw-quiubo` is blocked by the new plugin code-safety scanner on `openclaw-quiubo@2.6.69`, but the flagged patterns appear to be false positives from bundled/runtime code rather than malicious behavior.
## What happened
Running:
```bash
openclaw plugins update openclaw-quiubo
`... | closed | completed | true | 1 | [] | [] | 2026-04-04T21:23:34Z | 2026-04-28T05:09:47Z | 2026-04-25T13:52:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | albertovasquez | 216,257 | MDQ6VXNlcjIxNjI1Nw== | User | false |
openclaw/openclaw | 4,205,847,361 | I_kwDOQb6kR876sCNB | 61,043 | https://github.com/openclaw/openclaw/issues/61043 | https://api.github.com/repos/openclaw/openclaw/issues/61043 | Feature request: auto-create agent on new channel account QR scan (openclaw-weixin) | ## 需求
当使用 `openclaw channels login --channel openclaw-weixin` 扫码登录一个新的微信机器人时，自动创建一个独立的 agent（包括 workspace、session store、人格配置），并将该 accountId 绑定到新 agent。
## 当前行为
必须先手动在 `openclaw.json` 中配置 `agents.list` 和 `bindings`，然后再扫码登录。流程繁琐，尤其是需要管理多个微信机器人时。
## 期望行为
1. 扫码登录新微信机器人时，自动创建一个新 agent
2. 自动生成独立 workspace（如 `~/.openclaw... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T21:30:29Z | 2026-04-28T05:09:47Z | 2026-04-25T13:52:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hchwork | 2,044,007 | MDQ6VXNlcjIwNDQwMDc= | User | false |
openclaw/openclaw | 4,205,904,557 | I_kwDOQb6kR876sQKt | 61,048 | https://github.com/openclaw/openclaw/issues/61048 | https://api.github.com/repos/openclaw/openclaw/issues/61048 | [Bug]: Claude CLI backend produces zero stdout when spawned as headless subprocess (launchd/systemd) | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
The `claude-cli` backend silently fails when the gateway runs under launchd (macOS) or likely systemd. The Claude CLI process starts, runs through initialization (verified via `~/.claude/debug/` logs), but produces **zero bytes** on stdout and ... | closed | completed | true | 1 | [] | [] | 2026-04-04T22:09:01Z | 2026-04-28T05:09:46Z | 2026-04-25T13:52:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jeremiahadler-svg | 262,598,621 | U_kgDOD6bv3Q | User | false |
openclaw/openclaw | 4,205,914,956 | I_kwDOQb6kR876sStM | 61,050 | https://github.com/openclaw/openclaw/issues/61050 | https://api.github.com/repos/openclaw/openclaw/issues/61050 | [Bug]: configure wizard "Anthropic Claude CLI" option produces unresolvable claude-cli/ model prefix | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
OpenClaw 2026.4.2 (d74a122). Running openclaw configure → Model → Anthropic → "Anthropic Claude CLI" sets agents.defaults.model.primary to claude-cli/claude-sonnet-4-6 and migrates the allowlist. However:
No claude-cli auth... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-04-04T22:16:39Z | 2026-04-28T05:09:45Z | 2026-04-25T13:53:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | soderstrommikael-ux | 265,464,519 | U_kgDOD9Kqxw | User | false |
openclaw/openclaw | 4,205,940,662 | I_kwDOQb6kR876sY-2 | 61,055 | https://github.com/openclaw/openclaw/issues/61055 | https://api.github.com/repos/openclaw/openclaw/issues/61055 | CLI backend: null bytes in system prompt crash child_process.spawn | ## Bug
When using the `claude-cli` backend (`cliBackends` config), the system prompt passed via `--append-system-prompt` sometimes contains null bytes (`\0`). Node.js rejects these in `child_process.spawn` args:
```
TypeError [ERR_INVALID_ARG_VALUE]: The argument 'args[11]' must be a string without null bytes.
Receiv... | closed | completed | true | 1 | [] | [] | 2026-04-04T22:35:45Z | 2026-04-28T05:09:44Z | 2026-04-25T13:53:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | MrCalebH | 200,527,564 | U_kgDOC_POzA | User | false |
openclaw/openclaw | 4,205,972,803 | I_kwDOQb6kR876sg1D | 61,066 | https://github.com/openclaw/openclaw/issues/61066 | https://api.github.com/repos/openclaw/openclaw/issues/61066 | [Bug]: 2026.4.4 - Build fails at pnpm build:docker — parse error in extensions/anthropic/register.runtime.ts:388 | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
I am rebuilding my Docker container using latest build.
Version: 2026.4.4
Error: Build fails at pnpm build:docker — parse error in extensions/anthropic/register.runtime.ts:388
Bundler: rolldown (via tsdown)
### Steps to repr... | closed | not_planned | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-04T23:00:02Z | 2026-04-28T05:09:43Z | 2026-04-25T13:53:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | stephenredmond-straiteis | 244,375,177 | U_kgDODpDeiQ | User | false |
openclaw/openclaw | 4,206,008,847 | I_kwDOQb6kR876spoP | 61,074 | https://github.com/openclaw/openclaw/issues/61074 | https://api.github.com/repos/openclaw/openclaw/issues/61074 | openclaw-weixin plugin: inbound metadata missing channel=weixin, causing identity/persona not to load | ## Bug Description
**Plugin**: `openclaw-weixin` (built-in)
**OpenClaw Version**: 2026.4.2
### Problem
When messages arrive via the WeChat plugin, the **inbound metadata does not include `channel: "weixin"`** (or any channel identifier that matches the weixin plugin). This causes persona/identity routing based on inb... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T23:28:17Z | 2026-04-28T05:09:43Z | 2026-04-25T13:53:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rookiestar | 1,826,999 | MDQ6VXNlcjE4MjY5OTk= | User | false |
openclaw/openclaw | 4,206,026,085 | I_kwDOQb6kR876st1l | 61,079 | https://github.com/openclaw/openclaw/issues/61079 | https://api.github.com/repos/openclaw/openclaw/issues/61079 | [Bug]: MCP plugin streamable-http transport fails to connect to MCP servers | ## Bug: MCP plugin streamable-http transport fails to connect to MCP servers
### Bug Description
The OpenClaw MCP plugin (`openclaw-mcp-plugin`) fails to connect to MCP servers via streamable-http transport. Both jcodemunch-mcp and FastMCP servers fail with "Missing session ID" or similar session-related errors.
###... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T23:41:40Z | 2026-04-28T05:09:42Z | 2026-04-25T13:53:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Xurkon | 36,556,990 | MDQ6VXNlcjM2NTU2OTkw | User | false |
openclaw/openclaw | 4,206,039,476 | I_kwDOQb6kR876sxG0 | 61,083 | https://github.com/openclaw/openclaw/issues/61083 | https://api.github.com/repos/openclaw/openclaw/issues/61083 | Showcase: Zero-Cost Chinese Novel Writing + Multi-Scale News Briefing Skills | Hi OpenClaw team! 👋
I built and published 4 skills on ClawHub that I think could be great for the community.
## 📚 Novel Writer Suite (3 skills)
A complete zero-cost toolchain for writing Chinese web novels. From outline to 700k+ words, all on free API tiers.
- `novel-outliner` — Parse outlines → chapter prompts (t... | closed | not_planned | true | 1 | [] | [] | 2026-04-04T23:52:08Z | 2026-04-28T05:09:41Z | 2026-04-25T13:53:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Shine8592 | 126,399,105 | U_kgDOB4iygQ | User | false |
openclaw/openclaw | 4,206,076,487 | I_kwDOQb6kR876s6JH | 61,096 | https://github.com/openclaw/openclaw/issues/61096 | https://api.github.com/repos/openclaw/openclaw/issues/61096 | /model command fails when primary provider is rate-limited — override never persists | ## Bug
The `/model` slash command (Discord) and `//model` text directive fail to switch models when the primary provider is rate-limited or has a billing issue. The command reports success but the model doesn't actually change.
## Root Cause
Five separate allowlist validation gates prevent `/model` from persisting t... | closed | completed | true | 2 | [] | [] | 2026-04-05T00:19:47Z | 2026-04-28T05:09:40Z | 2026-04-25T13:53:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Tgk-30 | 129,732,092 | U_kgDOB7uN_A | User | false |
openclaw/openclaw | 4,206,077,341 | I_kwDOQb6kR876s6Wd | 61,097 | https://github.com/openclaw/openclaw/issues/61097 | https://api.github.com/repos/openclaw/openclaw/issues/61097 | [Bug] task_runs.requester_session_key NOT NULL constraint causes Gateway crash every ~30 minutes | ## Bug Description
Gateway crashes every ~30 minutes with SQLite constraint violation: `NOT NULL constraint failed: task_runs.requester_session_key`
## Affected Versions
- v2026.3.30 (first discovered)
- v2026.4.4 (still affected after upgrade)
## Error
```
[openclaw] Unhandled promise rejection: Error: NOT NULL con... | closed | completed | true | 1 | [] | [] | 2026-04-05T00:20:27Z | 2026-04-28T05:09:40Z | 2026-04-25T13:53:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gaojinha | 50,872,282 | MDQ6VXNlcjUwODcyMjgy | User | false |
openclaw/openclaw | 4,206,141,259 | I_kwDOQb6kR876tJ9L | 61,110 | https://github.com/openclaw/openclaw/issues/61110 | https://api.github.com/repos/openclaw/openclaw/issues/61110 | Bug: Provider errors incorrectly mapped to deactivated_workspace | # OpenClaw Bug Report: Provider Errors Incorrectly Mapped to "deactivated_workspace"
## Summary
When OpenAI Codex provider encounters errors (rate limit, token expiration), the error is incorrectly mapped and reported as `{"detail":{"code":"deactivated_workspace"}}` instead of the actual provider error.
## Environmen... | closed | completed | true | 1 | [] | [] | 2026-04-05T01:09:08Z | 2026-04-28T05:09:38Z | 2026-04-25T13:54:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | phheng | 44,972,754 | MDQ6VXNlcjQ0OTcyNzU0 | User | false |
openclaw/openclaw | 4,206,153,799 | I_kwDOQb6kR876tNBH | 61,116 | https://github.com/openclaw/openclaw/issues/61116 | https://api.github.com/repos/openclaw/openclaw/issues/61116 | acpx plugin: gateway passes --format as global flag, causing 'unknown option --format' error on ACP session spawn | ## Bug Report
**OpenClaw version:** 2026.4.2
**acpx version:** bundled 0.4.0 (latest on npm: 0.4.1)
**Claude Code version:** 2.1.92
**OS:** macOS 26.3.1 (arm64)
## Description
When the gateway attempts to spawn or probe an ACP session via the acpx backend, it fails with:
```
acpx ensureSession probing named session... | closed | completed | true | 1 | [] | [] | 2026-04-05T01:19:13Z | 2026-04-28T05:09:37Z | 2026-04-25T13:54:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cassixdrey | 271,751,408 | U_kgDOEDKY8A | User | false |
openclaw/openclaw | 4,206,158,083 | I_kwDOQb6kR876tOED | 61,117 | https://github.com/openclaw/openclaw/issues/61117 | https://api.github.com/repos/openclaw/openclaw/issues/61117 | Gateway ignores per-agent workspace path — all agents load default workspace SOUL.md | ## Bug Summary
When using `openclaw agent --agent <non-default-agent>`, the gateway always loads workspace files (SOUL.md, IDENTITY.md, AGENTS.md) from the **default agent's workspace** instead of the target agent's configured workspace. This causes all agents to share the default agent's identity.
## Environment
- ... | closed | completed | true | 2 | [] | [] | 2026-04-05T01:22:46Z | 2026-04-28T05:09:37Z | 2026-04-25T13:54:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rolandpg | 48,327,651 | MDQ6VXNlcjQ4MzI3NjUx | User | false |
openclaw/openclaw | 4,206,177,765 | I_kwDOQb6kR876tS3l | 61,124 | https://github.com/openclaw/openclaw/issues/61124 | https://api.github.com/repos/openclaw/openclaw/issues/61124 | Discord reconnect-exhausted error crashes gateway during health-monitor restart | ## Summary
The health monitor's stale-socket restart of the Discord channel causes an uncaught exception that crashes the entire gateway process.
## Repro
1. Run the gateway with Discord channel enabled
2. Wait for the health monitor to detect a stale socket on the Discord channel (~35 min with default timing)
3. He... | closed | completed | true | 1 | [] | [] | 2026-04-05T01:39:10Z | 2026-04-28T05:09:35Z | 2026-04-25T13:54:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | conradchu | 56,970 | MDQ6VXNlcjU2OTcw | User | false |
openclaw/openclaw | 4,206,182,162 | I_kwDOQb6kR876tT8S | 61,125 | https://github.com/openclaw/openclaw/issues/61125 | https://api.github.com/repos/openclaw/openclaw/issues/61125 | Feature Request: Multi-tenant / Multi-agent support on a single gateway | ## Motivation
Users who share a single machine (family members, small teams) want to run separate OpenClaw instances with isolated workspaces, memories, and identities — without managing multiple gateways, VMs, or requiring technical knowledge.
## Proposal
Add multi-tenant configuration to the gateway, allowing mult... | closed | completed | true | 1 | [] | [] | 2026-04-05T01:42:30Z | 2026-04-28T05:09:34Z | 2026-04-25T13:54:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | giroclaw | 272,048,226 | U_kgDOEDcgYg | User | false |
openclaw/openclaw | 4,206,242,944 | I_kwDOQb6kR876tiyA | 61,139 | https://github.com/openclaw/openclaw/issues/61139 | https://api.github.com/repos/openclaw/openclaw/issues/61139 | [Bug]: Tools (read/edit/exec) hang indefinitely on 2026.4.2 with k2p5 — downgrading to 2026.4.1 fixes it | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading to **OpenClaw 2026.4.2**, core tools (`read`, `edit`, `exec`) hang indefinitely after invocation. The tool request itself is sent quickly, but the **result channel never returns**, causing the agent to stall ... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-05T02:14:32Z | 2026-04-28T05:09:32Z | 2026-04-25T13:55:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | alvinwangtw | 273,666,823 | U_kgDOEE_TBw | User | false |
openclaw/openclaw | 4,206,250,322 | I_kwDOQb6kR876tklS | 61,142 | https://github.com/openclaw/openclaw/issues/61142 | https://api.github.com/repos/openclaw/openclaw/issues/61142 | [Bug]: Telegram photo captions arrive as [object Object] | ## Bug Description
Telegram messages sent as **photo with caption** arrive to the LLM as `[object Object]` instead of the caption text. Pure text messages and photos without captions work fine.
**OpenClaw Version:** 2026.4.2 (d74a122)
**OS:** macOS arm64 (Apple M4 Pro)
**Node:** v25.6.0
**Channel:** Telegram (direct ... | closed | completed | true | 1 | [] | [] | 2026-04-05T02:20:23Z | 2026-04-28T05:09:31Z | 2026-04-25T13:55:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | chappg | 5,059,215 | MDQ6VXNlcjUwNTkyMTU= | User | false |
openclaw/openclaw | 4,206,267,950 | I_kwDOQb6kR876to4u | 61,149 | https://github.com/openclaw/openclaw/issues/61149 | https://api.github.com/repos/openclaw/openclaw/issues/61149 | [Bug]: minimax image generation fails with error 2049 (invalid api key) — wrong API endpoint | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The `image_generate` tool fails when using `minimax/image-01` with error:
> MiniMax image generation API error (2049): invalid api key
### Steps to reproduce
1. Add MINIMAX_API_KEY to env vars
2. Set `agen... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T02:34:43Z | 2026-04-28T05:09:31Z | 2026-04-25T13:55:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jessicamullen | 46,801,715 | MDQ6VXNlcjQ2ODAxNzE1 | User | false |
openclaw/openclaw | 4,206,272,210 | I_kwDOQb6kR876tp7S | 61,152 | https://github.com/openclaw/openclaw/issues/61152 | https://api.github.com/repos/openclaw/openclaw/issues/61152 | [Feature]: add a provider-free localCommand cron payload for deterministic host jobs | ### Summary
OpenClaw cron currently supports `systemEvent` and `agentTurn`, but it has no provider-free execution style for deterministic host-local chores. This forces users to wrap tiny local jobs in isolated LLM runs. In practice that means provider latency/failures can cause false cron timeouts for work that is ot... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-05T02:38:06Z | 2026-04-28T05:09:30Z | 2026-04-25T13:55:34Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | gaofeng21cn | 5,626,387 | MDQ6VXNlcjU2MjYzODc= | User | false |
openclaw/openclaw | 4,206,291,149 | I_kwDOQb6kR876tujN | 61,157 | https://github.com/openclaw/openclaw/issues/61157 | https://api.github.com/repos/openclaw/openclaw/issues/61157 | [Bug]: before_tool_call hook does not intercept message tool calls | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Plugin before_tool_call hook does not intercept message tool calls, allowing non-Korean characters to bypass language filter
### Steps to reproduce
1. Install a plugin that registers before_tool_call with too... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T02:54:16Z | 2026-04-28T05:09:29Z | 2026-04-25T13:55:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kwj8677 | 54,247,509 | MDQ6VXNlcjU0MjQ3NTA5 | User | false |
openclaw/openclaw | 4,206,292,355 | I_kwDOQb6kR876tu2D | 61,158 | https://github.com/openclaw/openclaw/issues/61158 | https://api.github.com/repos/openclaw/openclaw/issues/61158 | Feature Request: Support sending voice messages in openclaw-weixin | ## Feature Description
The openclaw-weixin plugin currently supports receiving voice messages (automatically transcribed to text), but does not support sending voice messages.
## Use Case
When a user requests a voice reply (e.g., "send me a voice message"), the Bot needs to be able to send voice messages. Currently ... | closed | not_planned | true | 1 | [] | [] | 2026-04-05T02:54:59Z | 2026-04-28T05:09:28Z | 2026-04-25T13:55:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | longcatmao | 194,860,338 | U_kgDOC51VMg | User | false |
openclaw/openclaw | 4,206,320,697 | I_kwDOQb6kR876t1w5 | 61,163 | https://github.com/openclaw/openclaw/issues/61163 | https://api.github.com/repos/openclaw/openclaw/issues/61163 | [Skill] DashPass — AI-first credential management on Dash Platform | ### What
DashPass is an OpenClaw skill that provides AI-agent-friendly credential management using Dash Platform as a decentralized encrypted data store.
AI agents need API keys, tokens, WIFs, and other secrets — programmatically, at speed, with audit trails. DashPass stores credentials on Dash Platform (client-side ... | closed | not_planned | true | 3 | [] | [] | 2026-04-05T03:10:18Z | 2026-04-28T05:09:26Z | 2026-04-25T13:56:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | DashBot-0001 | 259,516,018 | U_kgDOD3fmcg | User | false |
openclaw/openclaw | 4,206,325,755 | I_kwDOQb6kR876t2_7 | 61,165 | https://github.com/openclaw/openclaw/issues/61165 | https://api.github.com/repos/openclaw/openclaw/issues/61165 | [Bug]: ACP dead session repair: unrecoverable retry loop + duplicate Discord thread messages | ## Description
When an ACP session dies (status: dead, summary: "queue owner unavailable"), the `acpx ensureSession` repair mechanism has two related issues:
1. **Unrecoverable retry loop**: Repair repeatedly attempts to resume a session whose backend process has fully exited (`Resource not found`), never giving up u... | closed | completed | true | 2 | [] | [] | 2026-04-05T03:12:58Z | 2026-04-28T05:09:25Z | 2026-04-25T13:56:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Masatsugar | 20,656,329 | MDQ6VXNlcjIwNjU2MzI5 | User | false |
openclaw/openclaw | 4,206,354,692 | I_kwDOQb6kR876t-EE | 61,170 | https://github.com/openclaw/openclaw/issues/61170 | https://api.github.com/repos/openclaw/openclaw/issues/61170 | [Bug] Subagent completion announcement permanently lost when channel is temporarily unreachable (QQ + upgrade scenario) | ## Bug description
I am an AI enthusiast, not a professional developer. I am reporting this issue with the help of an AI agent (Fairy) on behalf of myself as a user.
**Problem**: After upgrading OpenClaw (and reinstalling the QQ bot plugin due to a failed upgrade), subagent completion announcements stopped being deli... | closed | completed | true | 1 | [] | [] | 2026-04-05T03:33:40Z | 2026-04-28T05:09:25Z | 2026-04-25T13:56:16Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Lupin924 | 202,222,886 | U_kgDODA2tJg | User | false |
openclaw/openclaw | 4,206,360,940 | I_kwDOQb6kR876t_ls | 61,171 | https://github.com/openclaw/openclaw/issues/61171 | https://api.github.com/repos/openclaw/openclaw/issues/61171 | [Bug] cron rm command hangs indefinitely with no output or error |
## Bug Description
`openclaw cron rm <id>` hangs indefinitely with no output, even with valid job IDs. The command never returns and must be killed manually with Ctrl+C.
## Steps to Reproduce
1. Run `openclaw cron rm <valid-job-id>`
2. Command hangs with no output, no error, no prompt
3. Must use `Ctrl+C` to recover
... | closed | completed | true | 1 | [] | [] | 2026-04-05T03:38:11Z | 2026-04-28T05:09:24Z | 2026-04-25T13:56:23Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | karlhillx | 13,650,232 | MDQ6VXNlcjEzNjUwMjMy | User | false |
openclaw/openclaw | 4,206,379,800 | I_kwDOQb6kR876uEMY | 61,174 | https://github.com/openclaw/openclaw/issues/61174 | https://api.github.com/repos/openclaw/openclaw/issues/61174 | [Bug]: openclaw-weixin reply delivery fails after long agent turn — contextToken appears to expire | ## Bug Report
### Environment
- **OpenClaw version**: 2026.4.2
- **openclaw-weixin plugin version**: 2.1.6
- **OS**: Linux 6.1.0-41-amd64 (Debian)
- **Model**: MiniMax M2.7 Highspeed (via api.minimaxi.com/v1, openai-completions API)
- **Node.js**: v22.22.0
### What happened?
When the agent executes long-running task... | closed | not_planned | true | 2 | [] | [] | 2026-04-05T03:50:09Z | 2026-04-28T05:09:23Z | 2026-04-25T13:56:50Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dorokuma | 18,098,088 | MDQ6VXNlcjE4MDk4MDg4 | User | false |
openclaw/openclaw | 4,206,402,793 | I_kwDOQb6kR876uJzp | 61,178 | https://github.com/openclaw/openclaw/issues/61178 | https://api.github.com/repos/openclaw/openclaw/issues/61178 | [Feature]:SkillPilot Integration: Pre-LLM Skill Routing Engine | ### Summary
I built **SkillPilot** - a universal skill router that routes agent skills **before** LLM inference, reducing latency from 1-5 seconds to <25ms.
**GitHub:** https://github.com/RealTapeL/SkillPilot
**npm:** @realtapel/skillpilot
### Problem to solve
Current OpenClaw puts all skill descriptions into the... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-05T04:04:32Z | 2026-04-28T05:09:22Z | 2026-04-25T13:56:57Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | RealTapeL | 167,164,141 | U_kgDOCfa47Q | User | false |
openclaw/openclaw | 4,206,416,076 | I_kwDOQb6kR876uNDM | 61,181 | https://github.com/openclaw/openclaw/issues/61181 | https://api.github.com/repos/openclaw/openclaw/issues/61181 | [Bug]: openclaw browser start --profile openclaw uses ~/.openclaw-openclaw instead of ~/.openclaw and fails gateway auth | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
openclaw browser start --profile openclaw appears to resolve the OpenClaw home/config path incorrectly on Linux.
### Steps to reproduce
Instead of using the normal config:
• ~/.openclaw/openclaw.json
it tries to use:
• ... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-05T04:12:44Z | 2026-04-28T05:09:21Z | 2026-04-25T13:57:04Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | eliujl | 8,711,788 | MDQ6VXNlcjg3MTE3ODg= | User | false |
openclaw/openclaw | 4,206,449,764 | I_kwDOQb6kR876uVRk | 61,186 | https://github.com/openclaw/openclaw/issues/61186 | https://api.github.com/repos/openclaw/openclaw/issues/61186 | Cron timer tick crash: ReferenceError: fs$1 is not defined (v2026.4.2) | ## Bug Description
The cron scheduler's timer tick function crashes intermittently with:
```
[diagnostic] lane task error: lane=cron durationMs=32 error="ReferenceError: fs$1 is not defined"
```
And in the full gateway log:
```
{"err":"ReferenceError: fs$1 is not defined"} "cron: timer tick failed"
```
Source: `gat... | closed | completed | true | 1 | [] | [] | 2026-04-05T04:26:35Z | 2026-04-28T05:09:21Z | 2026-04-25T13:57:11Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cinapbot | 258,591,357 | U_kgDOD2nKfQ | User | false |
openclaw/openclaw | 4,206,464,177 | I_kwDOQb6kR876uYyx | 61,188 | https://github.com/openclaw/openclaw/issues/61188 | https://api.github.com/repos/openclaw/openclaw/issues/61188 | Feature request: tools.exec.obfuscationDetection config key to disable/raise command length threshold | ## Problem
The exec obfuscation detection has a hardcoded `MAX_COMMAND_CHARS = 10000` limit in `src/infra/exec-obfuscation-detect.ts`. When a command exceeds this length, `obfuscation.detected` is set to `true` and is ORed directly into `requiresAsk`, forcing an approval prompt **regardless** of `security: full` or `a... | closed | completed | true | 1 | [] | [] | 2026-04-05T04:32:31Z | 2026-04-28T05:09:20Z | 2026-04-25T13:57:18Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jarvisaibowen | 273,410,257 | U_kgDOEEvo0Q | User | false |
openclaw/openclaw | 4,206,507,020 | I_kwDOQb6kR876ujQM | 61,195 | https://github.com/openclaw/openclaw/issues/61195 | https://api.github.com/repos/openclaw/openclaw/issues/61195 | [Bug]: Telegram provider silently fails to start on fresh gateway boots in v2026.4.2 | ### Summary
Telegram provider silently fails to initialize on fresh gateway boots in v2026.4.2. No error is logged — the `[telegram] starting provider` line simply never appears. Gateway runs fine otherwise (agents, WebSocket, plugins all work). Rolling back to v2026.4.1 reliably fixes it.
### Reproduction
1. Instal... | closed | not_planned | true | 3 | [] | [] | 2026-04-05T04:54:47Z | 2026-04-28T05:09:19Z | 2026-04-25T13:57:25Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ghostai2026-jpg | 265,682,290 | U_kgDOD9X9cg | User | false |
openclaw/openclaw | 4,328,280,664 | I_kwDOQb6kR88AAAABAfxSWA | 71,609 | https://github.com/openclaw/openclaw/issues/71609 | https://api.github.com/repos/openclaw/openclaw/issues/71609 | Control UI device token mismatch loop after scope upgrade causes rate-limit lockout | ## Bug Description
After a device scope upgrade / device approval, the Control UI enters an authentication failure loop:
1. Browser holds stale local device credentials
2. Gateway switches to new device token/permission state after approval
3. Browser reconnects with mismatched token → `device_token_mismatch`
4. Fron... | closed | completed | true | 2 | [] | [] | 2026-04-25T13:57:28Z | 2026-05-01T04:43:53Z | 2026-04-28T10:38:56Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | ricksayhi | 279,263,232 | U_kgDOEKU4AA | User | false |
openclaw/openclaw | 4,206,612,758 | I_kwDOQb6kR876u9EW | 61,205 | https://github.com/openclaw/openclaw/issues/61205 | https://api.github.com/repos/openclaw/openclaw/issues/61205 | [Feature] Telegram-first OpenAI/Codex reconnect workflow with device-auth fallback | ## Summary
It would be very helpful if OpenClaw supported a first-class Telegram workflow for reconnecting OpenAI/Codex auth without requiring the user to log back into a VPS or remote shell every time auth needs to be refreshed.
The main goal is to reduce operational friction for Telegram users running OpenClaw on r... | closed | not_planned | true | 2 | [] | [] | 2026-04-05T05:40:39Z | 2026-04-28T05:09:18Z | 2026-04-25T13:57:39Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | RamaAditya49 | 213,913,142 | U_kgDODMAONg | User | false |
openclaw/openclaw | 4,206,648,097 | I_kwDOQb6kR876vFsh | 61,209 | https://github.com/openclaw/openclaw/issues/61209 | https://api.github.com/repos/openclaw/openclaw/issues/61209 | Ollama provider not registered in gateway runtime despite OLLAMA_API_KEY=ollama-local and implicit config | ## Summary
OpenClaw gateway on macOS/launchd still fails to register the Ollama provider at runtime even after:
- removing `models.providers.ollama` from `~/.openclaw/openclaw.json`
- setting `OLLAMA_API_KEY=ollama-local`
- restarting the gateway cleanly
Startup logs show the session/agent model as `ollama/qwen3.5:4b... | closed | completed | true | 1 | [] | [] | 2026-04-05T05:55:41Z | 2026-04-28T05:09:18Z | 2026-04-25T13:57:46Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rimkute | 112,832,114 | U_kgDOBrmucg | User | false |
openclaw/openclaw | 4,206,653,541 | I_kwDOQb6kR876vHBl | 61,210 | https://github.com/openclaw/openclaw/issues/61210 | https://api.github.com/repos/openclaw/openclaw/issues/61210 | Feature: Deduplicate repeated file reads in context | ## Problem
When an agent reads the same file multiple times within a session (e.g., repeatedly reading `package.json` or `AGENTS.md`), the full content is injected into context each time. This causes unnecessary context bloat, especially during long coding sessions.
Current mitigation: `contextPruning` soft-trims old... | closed | not_planned | true | 1 | [] | [] | 2026-04-05T05:58:02Z | 2026-04-28T05:09:17Z | 2026-04-25T13:57:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | PacoXXD | 100,558,748 | U_kgDOBf5nnA | User | false |
openclaw/openclaw | 4,206,760,679 | I_kwDOQb6kR876vhLn | 61,219 | https://github.com/openclaw/openclaw/issues/61219 | https://api.github.com/repos/openclaw/openclaw/issues/61219 | LINE channel fails to start after upgrading to v2026.4.2. Error: "Bundled plugin public surface access blocked for line via line/runtime-api.js: no bundled plugin manifest found for line" | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
1. npm install -g openclaw (upgrades to v2026.4.2)
2. LINE channel was working on v2026.3.28
3. After upgrade, gateway starts normally but LINE channel fails immediately and keeps restarting
### Steps to reproduce
## Envi... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-05T06:45:46Z | 2026-04-28T05:09:16Z | 2026-04-25T13:58:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | hsing930 | 104,485,097 | U_kgDOBjpQ6Q | User | false |
openclaw/openclaw | 4,206,791,675 | I_kwDOQb6kR876vov7 | 61,223 | https://github.com/openclaw/openclaw/issues/61223 | https://api.github.com/repos/openclaw/openclaw/issues/61223 | Ollama native API: thinking tokens cause silent message drop + client queue lockup | ## Summary
Two bugs conspire to make Ollama reasoning models completely non-functional in webchat. The first causes empty agent responses; the second causes all subsequent messages to be silently dropped.
## Environment
- OpenClaw version: 2026.4.2
- Provider: `ollama-ada` (custom Ollama instance on port 11435, nati... | closed | completed | true | 1 | [] | [] | 2026-04-05T07:04:27Z | 2026-04-28T05:09:15Z | 2026-04-25T13:58:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | L3G | 35,130,228 | MDQ6VXNlcjM1MTMwMjI4 | User | false |
openclaw/openclaw | 4,206,861,847 | I_kwDOQb6kR876v54X | 61,233 | https://github.com/openclaw/openclaw/issues/61233 | https://api.github.com/repos/openclaw/openclaw/issues/61233 | Bug: allow-always does not persist — each command triggers new approval | ## Bug Report: exec-approvals allow-always 失效
### Description
When a user clicks the approval prompt's **Allow Always** button, the approval is saved to exec-approvals.json with a hashed pattern instead of a stable command identifier. This causes subsequent executions of the same command to trigger a new approval pro... | closed | completed | true | 1 | [] | [] | 2026-04-05T07:38:46Z | 2026-04-28T05:09:15Z | 2026-04-25T13:58:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | GuanxiongWuEgo | 138,973,604 | U_kgDOCEiRpA | User | false |
openclaw/openclaw | 4,206,888,729 | I_kwDOQb6kR876wAcZ | 61,240 | https://github.com/openclaw/openclaw/issues/61240 | https://api.github.com/repos/openclaw/openclaw/issues/61240 | [Bug]: BREW NEEDS TO BE REMOVE AS THE MAIN INSTALLER | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Having homebrew as the main installer for skill is extremelly poor because it cannot run as root. This is an issue when you are working with lxcs or containers that you are keeping simple and not hard to manage with multiple... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-05T07:48:20Z | 2026-04-28T05:09:14Z | 2026-04-25T13:58:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fangmaster35 | 268,249,504 | U_kgDOD_0poA | User | false |
openclaw/openclaw | 4,206,922,662 | I_kwDOQb6kR876wIum | 61,241 | https://github.com/openclaw/openclaw/issues/61241 | https://api.github.com/repos/openclaw/openclaw/issues/61241 | [Bug]: Telegram replies can disappear after switching a direct session to xiaomi/mimo-v2-pro (reply generated, no sendMessage log) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
In a Telegram direct session using xiaomi/mimo-v2-pro, OpenClaw can generate assistant reply text internally but fail to deliver it to Telegram, with no corresponding sendMessage success or failure log for the ... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T08:00:46Z | 2026-04-28T05:09:13Z | 2026-04-25T13:58:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ha7rock | 33,951,467 | MDQ6VXNlcjMzOTUxNDY3 | User | false |
openclaw/openclaw | 4,206,980,952 | I_kwDOQb6kR876wW9Y | 61,252 | https://github.com/openclaw/openclaw/issues/61252 | https://api.github.com/repos/openclaw/openclaw/issues/61252 | [cli-backend] Codex companion plugin hijacks session ID → FailoverError after 2 turns | ## Environment
- OpenClaw version: 2026.4.2
- Claude CLI version: 2.1.89
- OS: macOS (arm64)
- CLI backend: `claude-cli` with `codex@openai-codex` plugin enabled in `~/.claude/settings.json`
## Bug: Codex companion plugin hijacks Claude CLI session ID → FailoverError after 2 turns
### Steps to Reproduce
1. Enable `... | closed | completed | true | 1 | [] | [] | 2026-04-05T08:40:06Z | 2026-04-28T05:09:12Z | 2026-04-25T13:58:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dabryant815 | 258,717,946 | U_kgDOD2u4-g | User | false |
openclaw/openclaw | 4,206,987,544 | I_kwDOQb6kR876wYkY | 61,255 | https://github.com/openclaw/openclaw/issues/61255 | https://api.github.com/repos/openclaw/openclaw/issues/61255 | [cli-backend] Conversations vanish on browser refresh — cli-backend does not persist messages to OpenClaw session store | ## Summary
When using `claude-cli/*` as the message provider, all conversations disappear after refreshing the webchat browser tab. Messages are only pushed via WebSocket in real-time but **never written to OpenClaw's session JSONL file**, so they are lost on any page reload.
## Environment
- OpenClaw version: 2026.... | closed | completed | true | 2 | [] | [] | 2026-04-05T08:44:02Z | 2026-04-28T05:09:12Z | 2026-04-25T13:59:03Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | dabryant815 | 258,717,946 | U_kgDOD2u4-g | User | false |
openclaw/openclaw | 4,207,006,500 | I_kwDOQb6kR876wdMk | 61,259 | https://github.com/openclaw/openclaw/issues/61259 | https://api.github.com/repos/openclaw/openclaw/issues/61259 | Bundled channel plugins cause stack overflow via jiti module loading | ## Summary
All bundled channel plugins fail to load with `RangeError: Maximum call stack size exceeded` during gateway startup. The stack overflow occurs in jiti's recursive module evaluator (`eval_evalModule` → `jitiRequire` chain), which creates a deeply nested call stack when loading compiled dist chunks. By the ti... | closed | completed | true | 1 | [] | [] | 2026-04-05T08:53:42Z | 2026-04-28T05:09:11Z | 2026-04-25T13:59:10Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | fjventura20 | 144,898,432 | U_kgDOCKL5gA | User | false |
openclaw/openclaw | 4,207,024,384 | I_kwDOQb6kR876whkA | 61,263 | https://github.com/openclaw/openclaw/issues/61263 | https://api.github.com/repos/openclaw/openclaw/issues/61263 | الانضمام إلى القائمة المشتركة الخاصة بي في Microsoft To Do | يُرجى الانضمام إلى قائمة بدون عنوان (1)، القائمة المشتركة الخاصة بي في Microsoft To Do.
https://to-do.microsoft.com/sharing?InvitationToken=2XjCR9Dwyawkc_1EqyKUIwN004phbqIMrTBFcE8zzXkxDAnVyuB9yVOS8hJ_Bh_Zc | closed | not_planned | true | 1 | [] | [] | 2026-04-05T09:05:48Z | 2026-04-28T05:09:10Z | 2026-04-25T13:59:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | almassolarenrgi | 170,923,824 | U_kgDOCjAXMA | User | false |
openclaw/openclaw | 4,207,153,462 | I_kwDOQb6kR876xBE2 | 61,277 | https://github.com/openclaw/openclaw/issues/61277 | https://api.github.com/repos/openclaw/openclaw/issues/61277 | [Bug]: App status shows 'idle' even when agent is actively running tasks | ## Bug Description
The app status bar always shows 'idle' even when the agent is actively executing tasks.
## Expected Behavior
When the agent is running a task (e.g., searching literature, generating content), the app should display 'running' status.
## Current Situation
- task_checkpoint.json is being updated c... | closed | completed | true | 1 | [] | [] | 2026-04-05T09:57:01Z | 2026-04-28T05:09:09Z | 2026-04-25T13:59:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | drzeast-png | 264,975,099 | U_kgDOD8sy-w | User | false |
openclaw/openclaw | 4,207,178,466 | I_kwDOQb6kR876xHLi | 61,283 | https://github.com/openclaw/openclaw/issues/61283 | https://api.github.com/repos/openclaw/openclaw/issues/61283 | [Bug] openclaw plugins doctor hangs indefinitely due to persistent plugin services | ## Bug Description
Running `openclaw plugins doctor` causes the command to hang indefinitely and never exit.
## Root Cause
The `doctor` command connects to the running OpenClaw gateway, which triggers full plugin initialization. Plugins like `memos-local` start persistent background services (e.g., an HTTP Viewer se... | closed | completed | true | 1 | [] | [] | 2026-04-05T10:14:31Z | 2026-04-28T05:09:09Z | 2026-04-25T13:59:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | prospect1314521 | 135,647,789 | U_kgDOCBXSLQ | User | false |
openclaw/openclaw | 4,207,188,553 | I_kwDOQb6kR876xJpJ | 61,288 | https://github.com/openclaw/openclaw/issues/61288 | https://api.github.com/repos/openclaw/openclaw/issues/61288 | [Bug]: Token Usage Not Captured for Bailian API (Streaming Mode) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Token usage data is NOT captured when using Bailian API. All session files show usage: { input: 0, output: 0 }, resulting in empty Grafana dashboards.
### Steps to reproduce
1. Configure OpenClaw with Bailian API (qwen3.5-... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-05T10:21:48Z | 2026-04-28T05:09:08Z | 2026-04-25T13:59:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | weblily588 | 266,917,648 | U_kgDOD-jXEA | User | false |
openclaw/openclaw | 4,328,286,367 | I_kwDOQb6kR88AAAABAfxonw | 71,610 | https://github.com/openclaw/openclaw/issues/71610 | https://api.github.com/repos/openclaw/openclaw/issues/71610 | MCP Server Memory Leak - zombie processes accumulate | ## Environment
- OS: Linux 6.8.0
- Memory: 3.6GB
- OpenClaw: 2026.4.8
## Problem
When calling `mcp-tfs-query` tools, gateway forks node child processes. Each invocation creates a new process, but they are not properly cleaned up after exit, accumulating into zombie processes.
## Reproduction
```bash
ps aux | grep mcp... | closed | completed | true | 1 | [] | [] | 2026-04-25T13:59:46Z | 2026-04-28T04:46:36Z | 2026-04-25T19:35:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | winningzhb | 241,090,910 | U_kgDODl7BXg | User | false |
openclaw/openclaw | 4,207,209,317 | I_kwDOQb6kR876xOtl | 61,292 | https://github.com/openclaw/openclaw/issues/61292 | https://api.github.com/repos/openclaw/openclaw/issues/61292 | [Test] GitHub CLI Integration Test by 诗诗 | This is a test issue created via GitHub CLI to demonstrate integration with OpenClaw.
## Test Details
- Created by: 诗诗 (帝诗公司 AI Assistant)
- Purpose: Testing GitHub CLI commands
- Date: 2026-04-05
- Status: Test issue, will be closed after verification
## Commands Tested
1. gh auth status - Authentication check
2. gh... | closed | not_planned | true | 1 | [] | [] | 2026-04-05T10:36:40Z | 2026-04-28T05:09:07Z | 2026-04-25T14:00:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | silicun-ai | 273,761,607 | U_kgDOEFFFRw | User | false |
openclaw/openclaw | 4,207,264,792 | I_kwDOQb6kR876xcQY | 61,303 | https://github.com/openclaw/openclaw/issues/61303 | https://api.github.com/repos/openclaw/openclaw/issues/61303 | [Bug]:[Bug] webchat channel 工具调用结果无法返回 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
Yes
### Summary
标题：[Bug] webchat channel 工具调用结果无法返回
环境：
- OpenClaw 2026.4.2 (d74a122)
- Gateway mode: local, bind: loopback
- Channel: webchat (内置 UI)
- tools.exec: {host:"auto", security:"allowlist", ask:"off"}
现象：
- agent 通过 webchat 对话... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-05T11:12:36Z | 2026-04-28T05:09:06Z | 2026-04-25T14:00:07Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bryan4gooner-pixel | 267,809,892 | U_kgDOD_Z0ZA | User | false |
openclaw/openclaw | 4,207,283,620 | I_kwDOQb6kR876xg2k | 61,313 | https://github.com/openclaw/openclaw/issues/61313 | https://api.github.com/repos/openclaw/openclaw/issues/61313 | [Bug]: Remote node exec completely non-functional in 2026.4.2 — all routing paths blocked | ## Summary
All paths to execute commands on a remote node are broken in OpenClaw 2026.4.2. Related to locked issues #20669 and #11832, but filing a new report since the problem persists with no fix in sight.
## Environment
- **Gateway:** Ubuntu VPS (Linux 6.8.0, x64), OpenClaw 2026.4.2 (d74a122)
- **Node:** Windows ... | closed | completed | true | 2 | [] | [] | 2026-04-05T11:23:04Z | 2026-04-28T05:09:06Z | 2026-04-25T14:00:14Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Re1ne99 | 126,322,360 | U_kgDOB4eGuA | User | false |
openclaw/openclaw | 4,207,345,542 | I_kwDOQb6kR876xv-G | 61,324 | https://github.com/openclaw/openclaw/issues/61324 | https://api.github.com/repos/openclaw/openclaw/issues/61324 | ACP thread-bound sessions deliver duplicate text on Discord (block reply not treated as visible) | ## Bug Description
Thread-bound ACP sessions on Discord deliver the same reply text twice within a single message. Every reply from the ACP session contains the content duplicated — the text appears once as a block reply and again as a final reply, concatenated into the same Discord message.
## Steps to Reproduce
1.... | closed | completed | true | 2 | [] | [] | 2026-04-05T11:57:11Z | 2026-04-28T05:09:05Z | 2026-04-25T14:00:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xiaoyuervae | 19,305,379 | MDQ6VXNlcjE5MzA1Mzc5 | User | false |
openclaw/openclaw | 4,207,372,071 | I_kwDOQb6kR876x2cn | 61,328 | https://github.com/openclaw/openclaw/issues/61328 | https://api.github.com/repos/openclaw/openclaw/issues/61328 | [Feature]: Failure-class aware provider fallback foundation | Since `openclaw/openclaw` currently does not have GitHub Discussions enabled, posting this here to scope-check a small runtime proposal before attempting any PR.
This is **not** a proposal for a product-layer routing platform, cost-governance UI, or broader “AI platform” abstraction. The scope is intentionally limited... | closed | completed | true | 1 | [] | [] | 2026-04-05T12:10:15Z | 2026-04-28T05:09:04Z | 2026-04-25T14:00:27Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | abcn1 | 200,472,276 | U_kgDOC_L21A | User | false |
openclaw/openclaw | 4,207,454,750 | I_kwDOQb6kR876yKoe | 61,336 | https://github.com/openclaw/openclaw/issues/61336 | https://api.github.com/repos/openclaw/openclaw/issues/61336 | [Bug] feishu_doc write action fails with 'Cannot read properties of undefined (reading 'split')' | ### Bug Description
The `feishu_doc` tool's `write` action fails with error: `Cannot read properties of undefined (reading 'split')`.
### Steps to Reproduce
1. Create a document (success):
```python
feishu_doc(action="create", title="Test Document")
# Returns: document_id: Mbv0drINRoC67qxqjxPctBMknqF
```
... | closed | completed | true | 1 | [] | [] | 2026-04-05T12:45:35Z | 2026-04-28T05:09:03Z | 2026-04-25T14:00:42Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | 89kpjddmtb-ui | 261,851,187 | U_kgDOD5uIMw | User | false |
openclaw/openclaw | 4,207,469,423 | I_kwDOQb6kR876yONv | 61,339 | https://github.com/openclaw/openclaw/issues/61339 | https://api.github.com/repos/openclaw/openclaw/issues/61339 | [CRITICAL BUG] Windows Task Scheduler Gateway Token Mismatch (1008) After Update to 2026.4.2 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Gateway token validation fails immediately after upgrading from 2026.4.1 to 2026.4.2. All OpenClaw CLI commands return "Gateway closed (1008): unauthorized: gateway token mismatch" error. Token is correctly configured but re... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-05T12:51:32Z | 2026-04-28T05:09:02Z | 2026-04-25T14:00:49Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | BurnoutMotors | 273,780,219 | U_kgDOEFGN-w | User | false |
openclaw/openclaw | 4,207,502,771 | I_kwDOQb6kR876yWWz | 61,348 | https://github.com/openclaw/openclaw/issues/61348 | https://api.github.com/repos/openclaw/openclaw/issues/61348 | Feature Request: Respect delivery.channel as exclusive target for cron jobs | **Problem**
When cron jobs are configured with `delivery.mode: "announce"` and a specific target channel (e.g. `delivery.channel: "telegram"`, `delivery.to: "<chat_id>"`), the result is broadcast to **all surfaces** — including the webchat TUI. This means cron job output appears as "System:" messages in the TUI even w... | closed | completed | true | 1 | [] | [] | 2026-04-05T13:07:06Z | 2026-04-28T05:09:02Z | 2026-04-25T14:00:56Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Banderasomatic | 31,406,708 | MDQ6VXNlcjMxNDA2NzA4 | User | false |
openclaw/openclaw | 4,207,562,227 | I_kwDOQb6kR876yk3z | 61,355 | https://github.com/openclaw/openclaw/issues/61355 | https://api.github.com/repos/openclaw/openclaw/issues/61355 | MemOS Viewer service.start() not invoked after plugin registration | ## Bug
MemOS Local plugin registers its Viewer service via `api.registerService()`, but the framework never calls the service's `start()` method. As a result, the Memory Viewer never starts on the configured port (default 18799).
## Evidence
Gateway startup logs show:
```
[plugins] memos-local: initialized (db: ~/.o... | closed | completed | true | 1 | [] | [] | 2026-04-05T13:36:10Z | 2026-04-28T05:09:01Z | 2026-04-25T14:01:17Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xpp0210 | 34,834,281 | MDQ6VXNlcjM0ODM0Mjgx | User | false |
openclaw/openclaw | 4,207,563,074 | I_kwDOQb6kR876ylFC | 61,356 | https://github.com/openclaw/openclaw/issues/61356 | https://api.github.com/repos/openclaw/openclaw/issues/61356 | [Bug]: WhatsApp: Watchdog timer not reset causes infinite reconnect loop + dedupe cache cleared causes duplicate replies | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
WhatsApp: Watchdog timer not reset causes infinite reconnect loop + dedupe cache cleared causes duplicate replies
### Steps to reproduce
NOT_ENOUGH_INFO
### Expected behavior
NOT_ENOUGH_INFO
### Actual beh... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T13:36:34Z | 2026-04-28T05:09:00Z | 2026-04-25T14:01:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bandyounghk | 265,701,424 | U_kgDOD9ZIMA | User | false |
openclaw/openclaw | 4,207,568,821 | I_kwDOQb6kR876yme1 | 61,359 | https://github.com/openclaw/openclaw/issues/61359 | https://api.github.com/repos/openclaw/openclaw/issues/61359 | [Bug]: Sub-agent model override not working - always falls back to MiniMax-M2.7 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When spawning sub-agents with `model` parameter (e.g., `model: "stepfun-plan/step-3.5-flash-2603"`), the agent ignores the model and uses MiniMax-M2.7 instead, despite `modelApplied: true` showing in the spawn ... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T13:39:06Z | 2026-04-28T05:09:00Z | 2026-04-25T14:01:31Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | MollaMert | 108,631,864 | U_kgDOBnmXOA | User | false |
openclaw/openclaw | 4,207,590,314 | I_kwDOQb6kR876yruq | 61,364 | https://github.com/openclaw/openclaw/issues/61364 | https://api.github.com/repos/openclaw/openclaw/issues/61364 | Gemini 2.5 Flash returns 400 via OpenClaw, works via direct curl | **Version:** OpenClaw 2026.3.27 (initially discovered), persists in 2026.4.4
**OS:** macOS (Darwin 25.4.0, arm64, M4 Mac mini)
**Node:** v25.8.2
## Summary
Gemini 2.5 Flash (`google/gemini-2.5-flash`) returns HTTP 400 errors when called through OpenClaw's model routing, but works correctly when called directly via `c... | closed | completed | true | 4 | [] | [] | 2026-04-05T13:50:38Z | 2026-04-28T05:08:59Z | 2026-04-25T14:01:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | FriDay026 | 273,780,127 | U_kgDOEFGNnw | User | false |
openclaw/openclaw | 4,207,630,865 | I_kwDOQb6kR876y1oR | 61,369 | https://github.com/openclaw/openclaw/issues/61369 | https://api.github.com/repos/openclaw/openclaw/issues/61369 | Feature: Native post-compaction workspace re-bootstrap still appears unresolved | ## Context
This is a follow-up to stale-closed issue #20225, which is now locked and cannot be commented on or revived in place.
## Why this needs a fresh issue
OpenClaw now has stronger workspace/bootstrap support than when #20225 was filed, including documented bootstrap injection and `agent:bootstrap` hooks. That i... | closed | completed | true | 1 | [] | [] | 2026-04-05T14:02:41Z | 2026-04-28T05:08:58Z | 2026-04-25T14:01:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mmartoccia | 4,631,577 | MDQ6VXNlcjQ2MzE1Nzc= | User | false |
openclaw/openclaw | 4,207,630,874 | I_kwDOQb6kR876y1oa | 61,370 | https://github.com/openclaw/openclaw/issues/61370 | https://api.github.com/repos/openclaw/openclaw/issues/61370 | bug(telegram): Compressed photo delivery still needs deterministic handling | ## Context
This is a follow-up to stale-closed issue #15078, which is now locked and cannot be commented on or revived in place.
## Why this needs a fresh issue
The previous issue was auto-closed for inactivity, but the issue history does not show a clear fix reference or closing PR. Because the old issue is locked, t... | closed | completed | true | 1 | [] | [] | 2026-04-05T14:02:42Z | 2026-04-28T05:08:57Z | 2026-04-25T14:01:52Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mmartoccia | 4,631,577 | MDQ6VXNlcjQ2MzE1Nzc= | User | false |
openclaw/openclaw | 4,207,651,992 | I_kwDOQb6kR876y6yY | 61,371 | https://github.com/openclaw/openclaw/issues/61371 | https://api.github.com/repos/openclaw/openclaw/issues/61371 | Feature Request: Silent/observe mode for WhatsApp groups — log messages with sender metadata without replying | ## Problem
I need to monitor WhatsApp group messages and store them (with sender name/number) in a database for analytics, without the agent ever replying in the group.
Currently there's no clean way to achieve this:
- **`requireMention: true`** — Agent doesn't reply in groups (good), but the `message:received` hook... | closed | completed | true | 2 | [] | [] | 2026-04-05T14:09:51Z | 2026-04-28T05:08:57Z | 2026-04-25T14:01:59Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | rarvizclawd | 273,651,226 | U_kgDOEE-WGg | User | false |
openclaw/openclaw | 4,207,692,856 | I_kwDOQb6kR876zEw4 | 61,376 | https://github.com/openclaw/openclaw/issues/61376 | https://api.github.com/repos/openclaw/openclaw/issues/61376 | createGoogleThinkingPayloadWrapper: strip store/service_tier/metadata only for google-generative-ai (collides with Codex subscription path) | ## Summary
In `createGoogleThinkingPayloadWrapper` (provider-stream), Google OpenAI-compat requests are dispatched with top-level `store`, `service_tier`, and `metadata` fields still attached. Google's OpenAI-compatible endpoint rejects these as unknown fields with HTTP 400. The wrapper already has a `model.api === "g... | closed | completed | true | 1 | [] | [] | 2026-04-05T14:28:40Z | 2026-04-28T05:08:56Z | 2026-04-25T14:02:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | jcat59 | 34,663,690 | MDQ6VXNlcjM0NjYzNjkw | User | false |
openclaw/openclaw | 4,207,755,982 | I_kwDOQb6kR876zULO | 61,384 | https://github.com/openclaw/openclaw/issues/61384 | https://api.github.com/repos/openclaw/openclaw/issues/61384 | [Bug]: Default memory convention and session-memory hook output are inconsistent (YYYY-MM-DD.md vs YYYY-MM-DD-slug.md) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
_(Thanks Krill for drafting the entire thing for me)_
With default AGENTS instructions + bundled session-memory hook enabled, /reset saves to memory/YYYY-MM-DD-slug.md but the next session startup still tries ... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T14:56:40Z | 2026-04-28T05:08:55Z | 2026-04-25T14:02:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | seemebreakthis | 5,940,900 | MDQ6VXNlcjU5NDA5MDA= | User | false |
openclaw/openclaw | 4,207,837,738 | I_kwDOQb6kR876zoIq | 61,391 | https://github.com/openclaw/openclaw/issues/61391 | https://api.github.com/repos/openclaw/openclaw/issues/61391 | Feature: Document and support intent-based model routing as a first-class pattern | ### Summary
OpenClaw users running multi-provider setups (e.g. Codex + Anthropic + xAI) would benefit from a documented pattern and optional built-in support for intent-based model routing — classifying each prompt and routing it to the cheapest appropriate model.
### Background
Filed originally as part of #61262 (n... | closed | completed | true | 4 | [] | [] | 2026-04-05T15:22:52Z | 2026-04-28T05:08:54Z | 2026-04-25T14:02:20Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cceausu | 171,296,923 | U_kgDOCjXImw | User | false |
openclaw/openclaw | 4,328,295,080 | I_kwDOQb6kR88AAAABAfyKqA | 71,613 | https://github.com/openclaw/openclaw/issues/71613 | https://api.github.com/repos/openclaw/openclaw/issues/71613 | Feature: channel reload <provider> — reinitialise single channel without full gateway restart | ## Summary
Currently, toggling `channels.<provider>.enabled` requires a full gateway restart. This is unnecessarily heavy for a single-channel state change and creates UX problems: all in-flight sessions pause, exec sessions are killed, and any tooling that triggers the restart (e.g. cron jobs) races against its own c... | closed | completed | true | 1 | [] | [] | 2026-04-25T14:03:03Z | 2026-04-28T04:46:35Z | 2026-04-25T19:35:24Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | axonrelaybot | 275,827,808 | U_kgDOEHDMYA | User | false |
openclaw/openclaw | 4,201,348,763 | I_kwDOQb6kR876a36b | 60,426 | https://github.com/openclaw/openclaw/issues/60426 | https://api.github.com/repos/openclaw/openclaw/issues/60426 | exec approval followup dispatch fails with gateway timeout when user does not respond within approval window | ## Bug
When an `exec` command requires approval (allowlist mode) and the user does not respond within the approval window (~32min), the subsequent async followup dispatch times out after 60 seconds, failing to deliver the agent's response to the user's chat channel.
## Evidence
From gateway logs (2026-04-04, user se... | closed | completed | true | 2 | [] | [] | 2026-04-03T16:51:42Z | 2026-04-28T05:08:53Z | 2026-04-25T14:04:38Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xiaolietu | 30,408,635 | MDQ6VXNlcjMwNDA4NjM1 | User | false |
openclaw/openclaw | 4,207,860,143 | I_kwDOQb6kR876ztmv | 61,395 | https://github.com/openclaw/openclaw/issues/61395 | https://api.github.com/repos/openclaw/openclaw/issues/61395 | [Bug]: lightContext not filtering workspace files (heartbeat + cron) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
## Issue Title
`lightContext: true` on heartbeat and cron payloads does not filter workspace files — full bootstrap context always injected
## Version
OpenClaw 2026.4.2 (d74a122)
## Description
Setting `light... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-05T15:29:06Z | 2026-04-28T05:08:53Z | 2026-04-25T14:04:45Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | drewclaw | 273,812,340 | U_kgDOEFILdA | User | false |
openclaw/openclaw | 4,208,106,429 | I_kwDOQb6kR8760pu9 | 61,439 | https://github.com/openclaw/openclaw/issues/61439 | https://api.github.com/repos/openclaw/openclaw/issues/61439 | bug(subagents): completion can be lost even when child run succeeded | ## Summary
A spawned subagent can complete successfully, but its completion handoff never arrives back in the parent conversation. The work is done, the child session history contains the result, but the parent thread never receives the completion event, causing the parent to think it is still waiting.
## What happene... | closed | completed | true | 1 | [] | [] | 2026-04-05T17:04:38Z | 2026-04-28T05:08:51Z | 2026-04-25T14:05:06Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | funmerlin | 258,679,497 | U_kgDOD2siyQ | User | false |
openclaw/openclaw | 4,208,114,730 | I_kwDOQb6kR8760rwq | 61,440 | https://github.com/openclaw/openclaw/issues/61440 | https://api.github.com/repos/openclaw/openclaw/issues/61440 | Gateway silently drops agent spawns when log file hits maxFileBytes cap | ## Summary
When the gateway log file reaches the `maxFileBytes` cap (default 512MB at `/tmp/openclaw/<date>.log`), the gateway emits `[openclaw] log file size cap reached; suppressing writes` and continues accepting WebSocket RPC calls (`chat.send`). However, agent sessions dispatched via `chat.send` during this state... | closed | not_planned | true | 1 | [] | [] | 2026-04-05T17:08:28Z | 2026-04-28T05:08:50Z | 2026-04-25T14:05:13Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mband0 | 31,228,859 | MDQ6VXNlcjMxMjI4ODU5 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.