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 | 4,219,668,357 | I_kwDOQb6kR877gweF | 62,655 | https://github.com/openclaw/openclaw/issues/62655 | https://api.github.com/repos/openclaw/openclaw/issues/62655 | openrouter: duplicate `auto` model in /models picker — bundled plugin id mismatch with pi-ai built-in | ## Summary
Running `/models` from any agent and selecting the `openrouter` provider shows two entries for the same underlying model:
- **OpenRouter Auto**
- **Auto Router**
Both buttons resolve to the same upstream call (`POST https://openrouter.ai/api/v1/chat/completions` with `model: openrouter/auto`), so this is ... | open | null | false | 2 | [] | [] | 2026-04-07T18:05:00Z | 2026-05-03T11:49:20Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | unitypeaceproject | 258,681,123 | U_kgDOD2spIw | User | false |
openclaw/openclaw | 4,338,718,429 | I_kwDOQb6kR88AAAABApuW3Q | 73,026 | https://github.com/openclaw/openclaw/issues/73026 | https://api.github.com/repos/openclaw/openclaw/issues/73026 | msteams: every inbound Bot Connector POST returns 401 after 4.24/4.25 runtime-deps move (CJS interop) | # msteams: every inbound Bot Connector POST returns 401 after 4.24/4.25 runtime-deps move (CJS interop)
## Summary
After upgrading to `2026.4.25` (and likely `2026.4.24` — the release that moved bundled deps to external runtime-deps), the bundled `msteams` plugin returns **HTTP 401** to every inbound `POST /api/messa... | closed | completed | true | 1 | [] | [] | 2026-04-27T20:35:34Z | 2026-04-30T04:33:05Z | 2026-04-27T20:54:25Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | kbrown10000 | 200,871,037 | U_kgDOC_kMfQ | User | false |
openclaw/openclaw | 4,338,638,195 | I_kwDOQb6kR88AAAABAppdcw | 73,017 | https://github.com/openclaw/openclaw/issues/73017 | https://api.github.com/repos/openclaw/openclaw/issues/73017 | [Bug]: Cron delivery schema rejects threadId field present in TypeScript types — type/API schema drift | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When passing `threadId` in a cron job's delivery object, the field is rejected at schema validation. ConfigDeliveryContextSchema (used elsewhere) does include threadId, but all three CronDelivery variant schemas do not — cre... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-27T20:19:09Z | 2026-04-30T04:33:06Z | 2026-04-27T20:42:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | coachsootz | 272,578,221 | U_kgDOED82rQ | User | false |
openclaw/openclaw | 4,338,676,915 | I_kwDOQb6kR88AAAABApr0sw | 73,021 | https://github.com/openclaw/openclaw/issues/73021 | https://api.github.com/repos/openclaw/openclaw/issues/73021 | Bug: Empty user message gets hallucinated content from DeepSeek, causing unauthorized config mutations | ### What happened
During a WebChat session, an **empty user message** (content: `""`) was presented to the assistant. The assistant (DeepSeek V4 Pro) then acted on specific text (`"putz den kleinen ram weg und aktivier tts"`) that never existed in the actual transcript. This caused the assistant to execute `openclaw c... | closed | completed | true | 2 | [] | [] | 2026-04-27T20:26:57Z | 2026-04-30T04:33:07Z | 2026-04-27T20:32:51Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | D3rPaPaH0d3n | 132,295,518 | U_kgDOB-KrXg | User | false |
openclaw/openclaw | 4,337,166,380 | I_kwDOQb6kR88AAAABAoPoLA | 72,905 | https://github.com/openclaw/openclaw/issues/72905 | https://api.github.com/repos/openclaw/openclaw/issues/72905 | [Bug]: 2026.4.25 Breaks Cloudflare AI Gateway | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Hello,
I upgraded to 2026.4.25 and using Anthropic models does not work anymore with the Cloudflare AI Gateway. Here's what shows up in the openclaw logs:
16:01:54+00:00 error diagnostic {"subsystem":"diagnostic"} lane tas... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-27T16:11:52Z | 2026-04-30T04:33:08Z | 2026-04-27T20:37:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | AaronFaby | 41,449,666 | MDQ6VXNlcjQxNDQ5NjY2 | User | false |
openclaw/openclaw | 4,338,643,849 | I_kwDOQb6kR88AAAABAppziQ | 73,018 | https://github.com/openclaw/openclaw/issues/73018 | https://api.github.com/repos/openclaw/openclaw/issues/73018 | [Bug]: Duplicated user/assistant turns in OpenClaw Control UI webchat | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Summary
While using OpenClaw Control UI in webchat, some user messages and assistant replies are duplicated in the conversation.
This does not appear to be only a rendering issue in the UI. The duplication also appears in ... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-27T20:20:08Z | 2026-04-30T04:33:09Z | 2026-04-27T20:35:07Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | uzikclaw | 279,827,681 | U_kgDOEK3U4Q | User | false |
openclaw/openclaw | 4,338,672,765 | I_kwDOQb6kR88AAAABAprkfQ | 73,020 | https://github.com/openclaw/openclaw/issues/73020 | https://api.github.com/repos/openclaw/openclaw/issues/73020 | Empty-response fallback leaks provider validation errors to users (self-propagating loop) | # Bug: Empty-response fallback path leaks provider validation errors to users
## Summary
When a primary model returns an empty response (0 output tokens) and the runtime falls over to a fallback provider that validates content blocks strictly (e.g., Anthropic on Bedrock), the fallback call fails with a validation err... | closed | completed | true | 1 | [] | [] | 2026-04-27T20:26:04Z | 2026-04-30T04:33:10Z | 2026-04-27T20:32:57Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | marcushale | 258,321,786 | U_kgDOD2Wteg | User | false |
openclaw/openclaw | 4,338,464,996 | I_kwDOQb6kR88AAAABApe45A | 73,006 | https://github.com/openclaw/openclaw/issues/73006 | https://api.github.com/repos/openclaw/openclaw/issues/73006 | [Bug]: `openclaw message send` CLI loads the full plugin registry on every call | ## Bug type
Bug (CLI scope defaults to "all", loads every plugin manifest just to send one message)
## Beta release blocker
No
## Summary
Sending a message via `openclaw message send` ends up loading every plugin shipped with openclaw before doing the send. With ~88 bundled plugins on this install that is 2-3 minu... | closed | completed | true | 1 | [] | [] | 2026-04-27T19:49:05Z | 2026-04-30T04:33:11Z | 2026-04-27T20:29:52Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jasonftl | 175,325,999 | U_kgDOCnNDLw | User | false |
openclaw/openclaw | 4,337,338,866 | I_kwDOQb6kR88AAAABAoaJ8g | 72,924 | https://github.com/openclaw/openclaw/issues/72924 | https://api.github.com/repos/openclaw/openclaw/issues/72924 | [Bug] Plugin runtime cache races between version-keyed and "unknown"-keyed extractors → repeated ENOTEMPTY on cold start | ## Summary
The gateway creates **two parallel plugin runtime cache directories** on every cold start, with mismatched extraction code paths that race and produce repeated `ENOTEMPTY` errors during plugin extraction. This is reproducible with **zero concurrent CLI activity** and a wiped cache.
Side effects observed:
-... | closed | completed | true | 3 | [] | [] | 2026-04-27T16:38:03Z | 2026-04-30T04:33:11Z | 2026-04-27T20:09:05Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | SledderBling | 279,591,126 | U_kgDOEKo41g | User | false |
openclaw/openclaw | 4,337,879,123 | I_kwDOQb6kR88AAAABAo7IUw | 72,971 | https://github.com/openclaw/openclaw/issues/72971 | https://api.github.com/repos/openclaw/openclaw/issues/72971 | [Bug]: Gateway startup blocks ~3 minutes between [hooks] loaded and [telegram] starting provider; OpenRouter/LiteLLM pricing fetches always hit 60s timeout despite endpoints being reachable in <200ms | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After every gateway start there is a ~3 minute window between [hooks] loaded and [telegram] [default] starting provider during which the agent is up but Telegram messages are not yet processed. The OpenRouter and LiteLLM mod... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-27T18:05:36Z | 2026-04-30T04:33:13Z | 2026-04-27T20:27:41Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | corruna | 74,627,018 | MDQ6VXNlcjc0NjI3MDE4 | User | false |
openclaw/openclaw | 4,337,153,594 | I_kwDOQb6kR88AAAABAoO2Og | 72,904 | https://github.com/openclaw/openclaw/issues/72904 | https://api.github.com/repos/openclaw/openclaw/issues/72904 | [Bug]: Gateway startup blocked by model pricing fetch timeout causing TUI connection failure | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
gateway connect failed: Error: gateway closed (1000)
### Steps to reproduce
主要内容：
- 问题：TUI 连接失败，显示 gateway closed (1000)
- 原因：Gateway 启动时被远程 pricing 请求（OpenRouter/LiteLLM）阻塞，每个请求超时 60 秒
- 影响：网络受限地区用户首次体验差
### Expected be... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-27T16:09:37Z | 2026-04-30T04:33:14Z | 2026-04-27T20:27:41Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | yhswwu | 263,410,287 | U_kgDOD7NSbw | User | false |
openclaw/openclaw | 4,336,924,227 | I_kwDOQb6kR88AAAABAoA2Qw | 72,880 | https://github.com/openclaw/openclaw/issues/72880 | https://api.github.com/repos/openclaw/openclaw/issues/72880 | Bug: pluginConfig not injected into hook handler event context for api.registerHook | ## Summary
When plugins register hooks via `api.registerHook()`, the `pluginConfig` from `openclaw.json` is **not** injected into the hook event context. Any plugin that accesses `ctx.pluginConfig` or `event.context.pluginConfig` inside a hook handler receives `undefined`.
## Affected Code
**File:** `src/plugins/reg... | closed | completed | true | 0 | [] | [] | 2026-04-27T15:32:13Z | 2026-04-30T04:33:16Z | 2026-04-27T20:19:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jalapeno777 | 3,149,646 | MDQ6VXNlcjMxNDk2NDY= | User | false |
openclaw/openclaw | 4,337,507,638 | I_kwDOQb6kR88AAAABAokdNg | 72,935 | https://github.com/openclaw/openclaw/issues/72935 | https://api.github.com/repos/openclaw/openclaw/issues/72935 | [Bug]: 2026.4.25 gateway becomes unresponsive and configured Telegram disappears from channels list | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
## Bug type
Regression (worked before, now fails)
## Beta release blocker
No
## Summary
After upgrading OpenClaw from `2026.4.24` to `2026.4.25`, the gateway became unresponsive to local WebSocket probes and `openclaw channels list ... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-27T17:05:15Z | 2026-04-30T04:33:19Z | 2026-04-27T20:12:42Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | firedigger | 4,905,460 | MDQ6VXNlcjQ5MDU0NjA= | User | false |
openclaw/openclaw | 4,337,208,075 | I_kwDOQb6kR88AAAABAoSLCw | 72,907 | https://github.com/openclaw/openclaw/issues/72907 | https://api.github.com/repos/openclaw/openclaw/issues/72907 | [Bug]: memory-core enabled but status reports Memory unavailable on v2026.4.25 | ## Summary
After upgrading a Linux VPS from OpenClaw `2026.4.24` to `2026.4.25`, `openclaw status` reports:
```text
Memory enabled (plugin memory-core) · unavailable
```
The gateway itself is reachable and Telegram works, but the memory status appears unavailable in the post-upgrade status report.
## Environment
-... | closed | completed | true | 3 | [] | [] | 2026-04-27T16:17:54Z | 2026-04-30T04:33:20Z | 2026-04-27T20:09:26Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | RamaAditya49 | 213,913,142 | U_kgDODMAONg | User | false |
openclaw/openclaw | 4,337,873,931 | I_kwDOQb6kR88AAAABAo60Cw | 72,970 | https://github.com/openclaw/openclaw/issues/72970 | https://api.github.com/repos/openclaw/openclaw/issues/72970 | Bug: QMD memory initialization fails after update - missing 'chokidar' in plugin-runtime-deps | ## Bug Report
### Description
After updating to OpenClaw **2026.4.25** (`openclaw-2026.4.25-babd02a671ff`), QMD memory startup fails for all agents with the following warning on every gateway start:
```
qmd memory startup initialization failed for agent "<agent-name>": Cannot find package 'chokidar' imported from /U... | closed | completed | true | 1 | [] | [] | 2026-04-27T18:04:39Z | 2026-04-30T04:33:22Z | 2026-04-27T20:08:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | colin-chang | 24,368,158 | MDQ6VXNlcjI0MzY4MTU4 | User | false |
openclaw/openclaw | 4,338,190,423 | I_kwDOQb6kR88AAAABApOIVw | 72,992 | https://github.com/openclaw/openclaw/issues/72992 | https://api.github.com/repos/openclaw/openclaw/issues/72992 | [Bug]: 2026.4.25: Gateway hangs/timeouts after update; plugin-runtime-deps missing packages | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
I am not very experienced with Linux/WSL, but several Discord users appear to be hitting the same issue, so I am collecting the symptoms here.
Multiple Discord users are reporting serious Gateway failures after updating to O... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-27T19:00:12Z | 2026-04-30T04:33:23Z | 2026-04-27T20:06:18Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Schnup03 | 145,503,562 | U_kgDOCKw1Sg | User | false |
openclaw/openclaw | 4,336,949,413 | I_kwDOQb6kR88AAAABAoCYpQ | 72,882 | https://github.com/openclaw/openclaw/issues/72882 | https://api.github.com/repos/openclaw/openclaw/issues/72882 | **[Bug]: 2026.4.25 — `chokidar` missing from plugin-runtime-deps bundle; `openclaw memory status` fails with ERR_MODULE_NOT_FOUND after upgrade** | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After upgrading from 2026.4.24 to 2026.4.25 via `openclaw update`, the `openclaw memory status` CLI command fails immediately with:
```
Error [ERR_MODULE_NOT_FOUND]: Cannot find package 'chokidar' imported from
~/.openclaw... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-27T15:36:35Z | 2026-04-30T04:33:24Z | 2026-04-27T20:06:10Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | shrimpy8 | 642,975 | MDQ6VXNlcjY0Mjk3NQ== | User | false |
openclaw/openclaw | 4,337,695,678 | I_kwDOQb6kR88AAAABAov7vg | 72,956 | https://github.com/openclaw/openclaw/issues/72956 | https://api.github.com/repos/openclaw/openclaw/issues/72956 | [Bug]: OpenClaw creates self-referential `openclaw-unknown-*` runtime dep cache that causes ENOTEMPTY crashes | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
After a clean `openclaw gateway` start, the `~/.openclaw/plugin-runtime-deps/` directory contains **two** cache directories:
```
openclaw-2026.4.24-da6bdffc3d96 ← correct, versioned
openclaw-unknown-5504045e5896 ← zombie... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-27T17:35:12Z | 2026-04-30T04:33:25Z | 2026-04-27T20:06:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jackkeller | 81,735 | MDQ6VXNlcjgxNzM1 | User | false |
openclaw/openclaw | 3,902,840,056 | I_kwDOQb6kR87ooJz4 | 9,795 | https://github.com/openclaw/openclaw/issues/9795 | https://api.github.com/repos/openclaw/openclaw/issues/9795 | [Security] sanitizeMimeType regex should be anchored and case-insensitive | ## Summary
In `src/media-understanding/apply.ts`, the `sanitizeMimeType` function's regex is not anchored at the end, which could allow malicious suffixes. Additionally, MIME types are case-insensitive per RFC standards.
## Impact
**Severity:** Medium (8/10)
**Type:** Security - Input validation
## Current Behavior
`... | closed | completed | true | 1 | [
"bug",
"security"
] | [] | 2026-02-05T17:58:47Z | 2026-04-30T04:33:26Z | 2026-04-27T20:02:32Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | ekson73 | 17,933,770 | MDQ6VXNlcjE3OTMzNzcw | User | false |
openclaw/openclaw | 4,338,364,401 | I_kwDOQb6kR88AAAABApYv8Q | 73,000 | https://github.com/openclaw/openclaw/issues/73000 | https://api.github.com/repos/openclaw/openclaw/issues/73000 | Session trajectory-path files accumulate indefinitely, causing node.list latency to degrade over time (20-35s observed at ~950 sessions) | ## Summary
Session trajectory-path files (`.trajectory-path.json`) are never pruned from the agent sessions directory. As the gateway accumulates sessions over days of operation, directory scan–dependent operations — particularly `node.list` and `session-write-lock` — degrade proportionally. At ~950 files we observe `... | closed | completed | true | 1 | [] | [] | 2026-04-27T19:30:56Z | 2026-04-30T04:33:27Z | 2026-04-27T20:00:54Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | jared-rebel | 147,876,109 | U_kgDOCNBpDQ | User | false |
openclaw/openclaw | 4,105,606,725 | I_kwDOQb6kR870tpZF | 50,826 | https://github.com/openclaw/openclaw/issues/50826 | https://api.github.com/repos/openclaw/openclaw/issues/50826 | Docs: alphabetical ordering violations in channel and provider lists | Several channel and provider lists are not in alphabetical order, violating the project guideline that docs lists should be ordered alphabetically.
## Instances
1. **docs/channels/index.md**: Slack (Sl) listed after Synology Chat (Sy) instead of before
2. **docs/providers/index.md**: Vercel AI Gateway listed before V... | closed | completed | true | 0 | [] | [] | 2026-03-20T04:08:25Z | 2026-04-30T04:33:27Z | 2026-04-27T20:00:06Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | scottgl9 | 10,623,288 | MDQ6VXNlcjEwNjIzMjg4 | User | false |
openclaw/openclaw | 4,337,789,017 | I_kwDOQb6kR88AAAABAo1oWQ | 72,962 | https://github.com/openclaw/openclaw/issues/72962 | https://api.github.com/repos/openclaw/openclaw/issues/72962 | [Bug]: `models status --json` can emit auth-profile logs to stdout before JSON in route-first mode |
## Summary
On OpenClaw `2026.4.25`, `openclaw models status --json` can emit a subsystem log line to stdout before the JSON payload. That makes stdout invalid JSON for automation that expects `--json` output to be parseable from byte 1.
Observed stdout shape:
```text
[agents/auth-profiles] read anthropic credentia... | closed | completed | true | 1 | [] | [] | 2026-04-27T17:50:00Z | 2026-04-30T04:33:28Z | 2026-04-27T19:52:32Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | vishutdhar | 68,405,187 | MDQ6VXNlcjY4NDA1MTg3 | User | false |
openclaw/openclaw | 4,338,398,580 | I_kwDOQb6kR88AAAABApa1dA | 73,004 | https://github.com/openclaw/openclaw/issues/73004 | https://api.github.com/repos/openclaw/openclaw/issues/73004 | Loopback Control UI WS preauth closes log as warn | ## Summary
When the gateway is running on loopback and the Control UI connects from the same origin, some browser websocket connections can close before auth completes and get logged as `WARN` even though they are benign local browser traffic.
This makes local startup and diagnostics look unhealthy even when the gatew... | closed | completed | true | 0 | [] | [] | 2026-04-27T19:36:55Z | 2026-04-30T04:33:29Z | 2026-04-27T19:46:35Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | LouisGameDev | 1,916,460 | MDQ6VXNlcjE5MTY0NjA= | User | false |
openclaw/openclaw | 4,338,169,255 | I_kwDOQb6kR88AAAABApM1pw | 72,989 | https://github.com/openclaw/openclaw/issues/72989 | https://api.github.com/repos/openclaw/openclaw/issues/72989 | `evaluateSessionFreshness` returns fresh-forever for future `updatedAt` timestamps (clock-skew edge case) | `evaluateSessionFreshness` (`src/config/sessions/reset-policy.ts:72-92`) returns `fresh: true` for any session whose `updatedAt` is in the future relative to `now`, regardless of reset mode. This silently defeats both `idle` and `daily` reset gates for any caller using the result.
## Code trace
```ts
// src/config/se... | closed | completed | true | 1 | [] | [] | 2026-04-27T18:56:22Z | 2026-04-30T04:33:31Z | 2026-04-27T19:31:18Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | martingarramon | 263,922,628 | U_kgDOD7sjxA | User | false |
openclaw/openclaw | 4,338,160,249 | I_kwDOQb6kR88AAAABApMSeQ | 72,986 | https://github.com/openclaw/openclaw/issues/72986 | https://api.github.com/repos/openclaw/openclaw/issues/72986 | [Bug]: web_search tool aborts silently with Gemini provider — API healthy, no err.log entry | ## 環境
- **OpenClaw 版本：** v2026.4.23
- **平台：** macOS arm64 (Apple Silicon)
## 問題描述
`web_search` 工具每次呼叫都回傳 `This operation was aborted`，完全無法使用。
### 確認 API 本身正常
直接用 curl 測試 Gemini API（同 endpoint / 同 API key），回應正常（<1s），排除上游 API 問題。
### 排除項目
- `gateway.err.log` 完全無 web_search 相關錯誤條目 — 異常發生但未被記錄
- 重啟 Gateway + `/new` ... | closed | not_planned | true | 1 | [] | [] | 2026-04-27T18:54:42Z | 2026-04-30T04:33:33Z | 2026-04-27T19:27:08Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | RoseKongPS | 270,016,663 | U_kgDOEBgglw | User | false |
openclaw/openclaw | 4,337,960,857 | I_kwDOQb6kR88AAAABApAHmQ | 72,977 | https://github.com/openclaw/openclaw/issues/72977 | https://api.github.com/repos/openclaw/openclaw/issues/72977 | Discord heartbeat DM targets degrade from user:<id> to channel:<id> | ## Summary
Discord heartbeat/session-derived DM routing can lose the `user:` prefix and degrade into `channel:<id>`, which causes outbound heartbeat delivery to fail with `Unknown Channel`.
## Environment
- OpenClaw version: `2026.4.25-beta.8`
- Repo HEAD observed: `ec71b01f71ad5156dd125fd019a9c59e29a3d53d`
- Host O... | closed | completed | true | 1 | [] | [] | 2026-04-27T18:20:01Z | 2026-04-30T04:33:34Z | 2026-04-27T19:24:59Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | LouisGameDev | 1,916,460 | MDQ6VXNlcjE5MTY0NjA= | User | false |
openclaw/openclaw | 4,338,273,462 | I_kwDOQb6kR88AAAABApTMtg | 72,998 | https://github.com/openclaw/openclaw/issues/72998 | https://api.github.com/repos/openclaw/openclaw/issues/72998 | Realtime voice calls remain active after caller hangup | ## Bug
Bundled realtime `voice-call` sessions can remain active after the caller hangs up. In the realtime handler, Twilio `stop` and WebSocket `close` currently close the realtime bridge, but the call is only finalized when the bridge closes with reason `error`. A normal hangup therefore does not emit `call.ended` an... | closed | completed | true | 0 | [] | [] | 2026-04-27T19:14:27Z | 2026-04-30T04:33:35Z | 2026-04-27T19:17:39Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | siegelz | 4,494,022 | MDQ6VXNlcjQ0OTQwMjI= | User | false |
openclaw/openclaw | 4,337,530,837 | I_kwDOQb6kR88AAAABAol31Q | 72,942 | https://github.com/openclaw/openclaw/issues/72942 | https://api.github.com/repos/openclaw/openclaw/issues/72942 | openclaw gateway stop is a no-op against foreground-launched gateway (no service-manager fallback) | > **In plain English:** if you start the gateway in foreground (e.g., `openclaw gateway` directly, no daemon wrapper), then run `openclaw gateway stop` to clean up, the stop command reports "Gateway service disabled" and exits success — but the running gateway process is still alive and the port is still bound. You hav... | closed | not_planned | true | 1 | [] | [] | 2026-04-27T17:09:09Z | 2026-04-30T04:33:38Z | 2026-04-27T17:29:24Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,337,524,054 | I_kwDOQb6kR88AAAABAoldVg | 72,939 | https://github.com/openclaw/openclaw/issues/72939 | https://api.github.com/repos/openclaw/openclaw/issues/72939 | Gateway crashes ~39ms after 'ready' on os.networkInterfaces() in restricted sandboxes (mDNS sidecar @homebridge/ciao) | > **In plain English:** when the gateway boots inside a sandbox that restricts `os.networkInterfaces()` (such as a NemoClaw / OpenShell pod), it crashes ~39ms after announcing "ready" because its mDNS sidecar (the `@homebridge/ciao` channel-discovery component) calls that syscall and gets a `SystemError`. The unhandled... | closed | not_planned | true | 1 | [] | [] | 2026-04-27T17:07:57Z | 2026-04-30T04:33:39Z | 2026-04-27T17:29:20Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mayank6136 | 16,468,779 | MDQ6VXNlcjE2NDY4Nzc5 | User | false |
openclaw/openclaw | 4,159,060,112 | I_kwDOQb6kR8735jiQ | 56,326 | https://github.com/openclaw/openclaw/issues/56326 | https://api.github.com/repos/openclaw/openclaw/issues/56326 | Bug: sessions_spawn exposes ACP-only fields and breaks runtime=subagent with schema-following models | ## Summary
`sessions_spawn` currently exposes ACP-only fields such as `streamTo` and `resumeSessionId` in the public tool schema, even when the intended runtime is `subagent`.
With schema-following models, this can cause subagent spawns to fail with errors like:
- `streamTo is only supported for runtime=acp; got run... | closed | completed | true | 4 | [
"clawsweeper"
] | [] | 2026-03-28T10:00:24Z | 2026-04-30T04:33:42Z | 2026-04-27T16:43:17Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | shenkq97 | 195,706,098 | U_kgDOC6o88g | User | false |
openclaw/openclaw | 4,337,071,146 | I_kwDOQb6kR88AAAABAoJ0Kg | 72,897 | https://github.com/openclaw/openclaw/issues/72897 | https://api.github.com/repos/openclaw/openclaw/issues/72897 | OpenAI Responses incremental context can replay orphan toolResult without matching toolCall | ## Summary
E2E test issue from IYENsystem/OpenClaw integration: OpenAI Responses incremental turns can fail when a stale/orphan `toolResult` is replayed without the matching assistant `toolCall` in the same incremental input window.
## Observed behavior
In long-running OpenClaw sessions with tool use + compaction/ru... | closed | not_planned | true | 2 | [] | [] | 2026-04-27T15:56:38Z | 2026-04-30T04:33:43Z | 2026-04-27T16:05:01Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | IYENTeam | 270,059,784 | U_kgDOEBjJCA | User | false |
openclaw/openclaw | 4,355,296,207 | I_kwDOQb6kR88AAAABA5iLzw | 74,845 | https://github.com/openclaw/openclaw/issues/74845 | https://api.github.com/repos/openclaw/openclaw/issues/74845 | [Feature]: Support long-running node-host commands with progress updates | ### Summary
Add first-class support for long-running node-host commands that can report progress and intermediate updates back to the caller.
### Problem to solve
Right now, node-host commands seem best suited for short-lived request/response work. That makes more persistent or interactive node-side tasks harder to ... | open | null | false | 1 | [
"enhancement"
] | [] | 2026-04-30T04:24:54Z | 2026-05-03T08:26:27Z | null | NONE | null | 20260503T090142Z | 2026-05-03T09:01:42Z | aehrt55 | 8,266,757 | MDQ6VXNlcjgyNjY3NTc= | User | false |
openclaw/openclaw | 4,355,315,989 | I_kwDOQb6kR88AAAABA5jZFQ | 74,848 | https://github.com/openclaw/openclaw/issues/74848 | https://api.github.com/repos/openclaw/openclaw/issues/74848 | [Bug]: macOS App node repeatedly disconnects with "cancelled" while CLI node works | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When running OpenClaw macOS App in remote mode, the app successfully connects the control channel, but the mac node repeatedly fails with:
mac node disconnected: connect failed: connect to gateway @ wss://oc.xxx.xyz: cancell... | open | null | false | 2 | [
"bug",
"regression"
] | [] | 2026-04-30T04:31:17Z | 2026-05-03T11:50:31Z | null | NONE | null | 20260503T120038Z | 2026-05-03T12:00:38Z | aiastia | 19,776,350 | MDQ6VXNlcjE5Nzc2MzUw | User | false |
openclaw/openclaw | 4,355,296,143 | I_kwDOQb6kR88AAAABA5iLjw | 74,844 | https://github.com/openclaw/openclaw/issues/74844 | https://api.github.com/repos/openclaw/openclaw/issues/74844 | MCP server tools missing from agent's request body in 2026.4.26 (gmail/outlook configured but not exposed to model) | # MCP server tools missing from agent's request body in 2026.4.26
## Summary
Configured MCP servers (`gmail`, `outlook`) spawn correctly and respond to direct stdio JSON-RPC probes with their full tool catalog, but their tools never appear in the `tools[]` array of the request body that the gateway sends to the LLM p... | closed | completed | true | 7 | [] | [] | 2026-04-30T04:24:52Z | 2026-05-03T04:33:40Z | 2026-04-30T15:20:21Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | mpalermiti | 253,352,132 | U_kgDODxnYxA | User | false |
openclaw/openclaw | 4,355,221,339 | I_kwDOQb6kR88AAAABA5dnWw | 74,814 | https://github.com/openclaw/openclaw/issues/74814 | https://api.github.com/repos/openclaw/openclaw/issues/74814 | Safety: steer messages must interrupt active tool chains at tool boundaries | ## Problem
Inbound steering currently appears unable to reliably interrupt long active tool chains at the next tool boundary. In a Signal DM configured with `messages.queue.mode = steer` and `byChannel.signal = steer`, new corrective messages were delivered only after the assistant completed a long sequence of tool ca... | closed | not_planned | true | 3 | [] | [] | 2026-04-30T04:00:12Z | 2026-05-02T04:26:01Z | 2026-04-30T04:04:56Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | howardclaw26 | 273,194,802 | U_kgDOEEifMg | User | false |
openclaw/openclaw | 4,355,355,543 | I_kwDOQb6kR88AAAABA5lzlw | 74,858 | https://github.com/openclaw/openclaw/issues/74858 | https://api.github.com/repos/openclaw/openclaw/issues/74858 | [Bug]: Telegram group messages silently dropped in 2026.4.27 (regression) | ---
name: Bug report
about: '[Bug]: Telegram group messages silently dropped in 2026.4.27 (regression)'
labels: 'bug,telegram,regression'
title: '[Bug]: Telegram group messages silently dropped in 2026.4.27 (regression)'
---
## Summary
In OpenClaw 2026.4.27, Telegram group messages are received by the gateway and pro... | closed | not_planned | true | 1 | [] | [] | 2026-04-30T04:43:08Z | 2026-05-03T04:38:41Z | 2026-04-30T04:47:24Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | mlaihk | 25,972,362 | MDQ6VXNlcjI1OTcyMzYy | User | false |
openclaw/openclaw | 4,355,372,962 | I_kwDOQb6kR88AAAABA5m3og | 74,859 | https://github.com/openclaw/openclaw/issues/74859 | https://api.github.com/repos/openclaw/openclaw/issues/74859 | [Bug]: abortable(activeSession.prompt()) creates zombie Agent loop when signal is pre-aborted | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When `params.abortSignal` is already aborted before `activeSession.prompt()` is called (e.g. rapid consecutive messages with `messages.queue.mode: "interrupt"`), `abortable()` immediately rejects but the `promp... | open | null | false | 1 | [] | [] | 2026-04-30T04:48:05Z | 2026-05-03T10:41:46Z | null | CONTRIBUTOR | null | 20260503T120038Z | 2026-05-03T12:00:38Z | zhumengzhu | 4,508,623 | MDQ6VXNlcjQ1MDg2MjM= | User | false |
openclaw/openclaw | 4,355,379,973 | I_kwDOQb6kR88AAAABA5nTBQ | 74,861 | https://github.com/openclaw/openclaw/issues/74861 | https://api.github.com/repos/openclaw/openclaw/issues/74861 | Redacted | Redacted at reporter request.
Please disregard this issue. If needed, a privacy-scrubbed report will be sent through a private channel.
| closed | completed | true | 1 | [] | [] | 2026-04-30T04:49:56Z | 2026-05-03T04:38:32Z | 2026-04-30T04:54:36Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | emperor-studio | 276,081,351 | U_kgDOEHSqxw | User | false |
openclaw/openclaw | 4,355,230,082 | I_kwDOQb6kR88AAAABA5eJgg | 74,818 | https://github.com/openclaw/openclaw/issues/74818 | https://api.github.com/repos/openclaw/openclaw/issues/74818 | [Bug]: 2026.4.27 plugin runtime-deps ENOENT race during hot channel reload leaves outbound permanently broken (regression of #55551) | ## Summary
On openclaw 2026.4.27 (macOS LaunchAgent gateway), any live edit to \`openclaw.json\` that triggers a channel hot-reload can fail plugin registration with \`ENOENT\` on a runtime-deps file **that demonstrably exists on disk milliseconds later**. When this hits the telegram plugin, it leaves the channel in a... | closed | completed | true | 4 | [] | [] | 2026-04-30T04:03:20Z | 2026-05-02T04:25:35Z | 2026-04-30T04:08:19Z | NONE | null | 20260502T060142Z | 2026-05-02T06:01:42Z | solosage1 | 158,075,162 | U_kgDOCWwJGg | User | false |
openclaw/openclaw | 4,355,384,752 | I_kwDOQb6kR88AAAABA5nlsA | 74,862 | https://github.com/openclaw/openclaw/issues/74862 | https://api.github.com/repos/openclaw/openclaw/issues/74862 | [Bug]: Session stuck in `processing` state causing infinite loop and API rate limiting | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
After a user message, the session gets stuck in `processing` state indefinitely; the model enters an infinite loop generating the same response, blocking the event loop (eventLoopUtilization=1) and triggering API rate limiti... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-30T04:51:20Z | 2026-05-03T04:38:28Z | 2026-04-30T04:56:16Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | asscan | 57,783,184 | MDQ6VXNlcjU3NzgzMTg0 | User | false |
openclaw/openclaw | 4,355,377,903 | I_kwDOQb6kR88AAAABA5nK7w | 74,860 | https://github.com/openclaw/openclaw/issues/74860 | https://api.github.com/repos/openclaw/openclaw/issues/74860 | Gateway 7-second event-loop stalls on 2026.4.27 — createOpenClawTools synchronously rebuilds plugin registry + provider-auth env-var resolution per tool-creation pass | ## Summary
After upgrading from `2026.4.15` → `2026.4.27`, the gateway exhibits ~7-second synchronous event-loop stalls during normal agent activity, blocking Telegram polling and forcing watchdog-triggered container restarts. 24h post-upgrade window: **27 gateway crashes, 25 Tier 2 (full container restart)**. Pre-upg... | closed | completed | true | 5 | [] | [] | 2026-04-30T04:49:23Z | 2026-05-04T04:36:34Z | 2026-05-01T11:48:06Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | mids-neo | 117,532,760 | U_kgDOBwFoWA | User | false |
openclaw/openclaw | 4,355,427,349 | I_kwDOQb6kR88AAAABA5qMFQ | 74,864 | https://github.com/openclaw/openclaw/issues/74864 | https://api.github.com/repos/openclaw/openclaw/issues/74864 | [Bug]: subagent-orphan-recovery resurrects a wedged subagent that immediately re-wedges the gateway, creating a restart-loop death spiral | ## Summary
On gateway restart, \`subagent-orphan-recovery\` resumes orphaned subagent sessions. If one of those sessions was wedged at the time the previous gateway died (e.g. event-loop-saturating prompt that never completes), the recovery resurrects it into the fresh gateway — which immediately re-wedges under the s... | closed | completed | true | 2 | [] | [] | 2026-04-30T05:02:59Z | 2026-05-03T04:37:19Z | 2026-04-30T13:57:23Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | solosage1 | 158,075,162 | U_kgDOCWwJGg | User | false |
openclaw/openclaw | 4,184,749,598 | I_kwDOQb6kR875bjYe | 58,818 | https://github.com/openclaw/openclaw/issues/58818 | https://api.github.com/repos/openclaw/openclaw/issues/58818 | Feature: guarantee last N raw messages in agent context (survive compaction and session reset) | ## Problem
When an agent wakes up in a session, its context window is built from the current session's JSONL transcript. Two mechanisms destroy or compress older messages:
1. **Daily reset** (default 4am): mints a new `sessionId` → the agent starts with zero messages. Previous conversation is gone from context entire... | open | null | false | 3 | [] | [] | 2026-04-01T07:19:16Z | 2026-05-01T04:43:20Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | atticusfawler | 268,220,059 | U_kgDOD_y2mw | User | false |
openclaw/openclaw | 4,355,061,032 | I_kwDOQb6kR88AAAABA5T1KA | 74,776 | https://github.com/openclaw/openclaw/issues/74776 | https://api.github.com/repos/openclaw/openclaw/issues/74776 | gateway: prewarmConfiguredPrimaryModel blocks channel startup for ~4 minutes on every restart | ## Summary
After every gateway restart, there is a **~4 minute gap** before channels (Feishu, Telegram, etc.) become ready, and the **first message** after restart takes 3–6 minutes to process. During message processing, all WebSocket/HTTP connections drop simultaneously.
After thorough investigation (tracing code pa... | closed | completed | true | 1 | [] | [] | 2026-04-30T03:14:05Z | 2026-05-03T04:38:26Z | 2026-04-30T05:14:42Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | Eric-CN-MS | 126,437,458 | U_kgDOB4lIUg | User | false |
openclaw/openclaw | 4,355,462,791 | I_kwDOQb6kR88AAAABA5sWhw | 74,868 | https://github.com/openclaw/openclaw/issues/74868 | https://api.github.com/repos/openclaw/openclaw/issues/74868 | bug: visibleReplies message_tool mode silently drops replies when message tool is unavailable | ## Summary
When `messages.groupChat.visibleReplies` is set to `"message_tool"` (the new default in 2026.4.27), group/channel replies are suppressed and the agent is expected to use `message(action=send)` to post visible output. However, **there is no guard to check whether the message tool is actually available** to t... | closed | completed | true | 2 | [] | [] | 2026-04-30T05:12:37Z | 2026-05-03T04:37:02Z | 2026-04-30T14:27:42Z | CONTRIBUTOR | null | 20260503T060030Z | 2026-05-03T06:00:30Z | kagura-agent | 268,167,063 | U_kgDOD_vnlw | User | false |
openclaw/openclaw | 4,355,448,561 | I_kwDOQb6kR88AAAABA5re8Q | 74,865 | https://github.com/openclaw/openclaw/issues/74865 | https://api.github.com/repos/openclaw/openclaw/issues/74865 | [Bug]: Regression: /subagents spawn fails (legacy context engine not initialized) | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Regression: /subagents spawn fails (legacy context engine not initialized).
### Steps to reproduce
/subagents spawn
### Expected behavior
Work with subagents.
### Actual behavior
Can't work with subagents.
### OpenCla... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-30T05:09:01Z | 2026-05-03T04:38:08Z | 2026-04-30T07:08:44Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | hategthb | 280,593,284 | U_kgDOELmDhA | User | false |
openclaw/openclaw | 4,164,381,818 | I_kwDOQb6kR874N2x6 | 57,067 | https://github.com/openclaw/openclaw/issues/57067 | https://api.github.com/repos/openclaw/openclaw/issues/57067 | Agents page: dropdown selector causes too much friction, should show agent list directly | ## Feature Request / UX Regression
**Summary:** The new Agents page design puts all agents behind a dropdown selector, causing significant friction compared to the previous design.
## Problems
**1. Agent list hidden in dropdown**
All agents are now hidden behind a single dropdown. Previously agents were visible at a... | open | null | false | 2 | [] | [] | 2026-03-29T13:47:02Z | 2026-05-01T06:04:25Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | fwends | 4,113,483 | MDQ6VXNlcjQxMTM0ODM= | User | false |
openclaw/openclaw | 4,153,871,295 | I_kwDOQb6kR873lwu_ | 55,878 | https://github.com/openclaw/openclaw/issues/55878 | https://api.github.com/repos/openclaw/openclaw/issues/55878 | [Bug]: macOS app crashes in WebSocketTaskBox.sendPing() and no stable menu bar icon appears | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Hi,
the macOS OpenClaw app installed via DMG starts, but no stable menu bar icon appears. The process stays alive briefly and then crashes, so chat cannot be opened from the menu bar.
### Steps to reproduce
OpenClaw proces... | closed | not_planned | true | 1 | [
"bug",
"regression"
] | [] | 2026-03-27T15:51:31Z | 2026-05-03T04:38:25Z | 2026-04-30T05:18:59Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | kkormesser | 27,818,245 | MDQ6VXNlcjI3ODE4MjQ1 | User | false |
openclaw/openclaw | 4,154,707,173 | I_kwDOQb6kR873o8zl | 55,942 | https://github.com/openclaw/openclaw/issues/55942 | https://api.github.com/repos/openclaw/openclaw/issues/55942 | [Bug]: NVIDIA NIM Kimi 2.5 tool execution completely broken in OpenClaw > 2026.2.25 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
# Regression: NVIDIA NIM Kimi 2.5 tool execution completely broken in OpenClaw > 2026.2.25
**Issue Type:** Bug (Regression)
**Affected Version:** OpenClaw > 2026.2.25 (2026.3.7–2026.3.24)
**Last Working Version:** 2026.... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-03-27T17:38:21Z | 2026-05-03T04:38:24Z | 2026-04-30T05:19:09Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | asasoft | 2,089,904 | MDQ6VXNlcjIwODk5MDQ= | User | false |
openclaw/openclaw | 4,176,192,954 | I_kwDOQb6kR87466W6 | 58,161 | https://github.com/openclaw/openclaw/issues/58161 | https://api.github.com/repos/openclaw/openclaw/issues/58161 | [Feature]: Support MCP elicitations in embedded Pi MCP client | ## Summary
Support [MCP elicitations](https://modelcontextprotocol.io/specification/draft/client/elicitation#capabilities) in the embedded Pi MCP client so plugin-backed MCP tools can request dynamic runtime approval from the user.
## Problem to solve
The embedded Pi MCP runtime currently supports calling MCP tools,... | open | null | false | 2 | [] | [] | 2026-03-31T06:18:11Z | 2026-05-01T06:04:33Z | null | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | kevinlin-openai | 177,054,202 | U_kgDOCo2h-g | User | false |
openclaw/openclaw | 4,163,454,127 | I_kwDOQb6kR874KUSv | 56,898 | https://github.com/openclaw/openclaw/issues/56898 | https://api.github.com/repos/openclaw/openclaw/issues/56898 | wacli skill: add recipient allowlist and confirm-before-send enforcement | ## Summary
The `wacli` skill (WhatsApp CLI wrapper) currently relies on soft agent-level guardrails in its SKILL.md. Since wacli can send messages to arbitrary WhatsApp contacts, stronger safety patterns should be enforced at the SKILL.md instruction level to prevent accidental or unauthorized message delivery.
## Ba... | open | null | false | 2 | [] | [] | 2026-03-29T09:00:40Z | 2026-05-01T06:04:22Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | xinbenlv | 640,325 | MDQ6VXNlcjY0MDMyNQ== | User | false |
openclaw/openclaw | 4,166,173,358 | I_kwDOQb6kR874UsKu | 57,355 | https://github.com/openclaw/openclaw/issues/57355 | https://api.github.com/repos/openclaw/openclaw/issues/57355 | Plugin SDK: add native registerMcpServer for managed MCP | ## Summary
Add a native `registerMcpServer` surface to the plugin SDK so plugins can register managed stdio MCP servers and inject them into embedded OpenAI-backed runs.
Why this matters:
- Some plugin features want to expose a dynamic tool/app inventory rather than a fixed tool list known at plugin load time.
- For e... | open | null | false | 2 | [] | [] | 2026-03-30T00:17:06Z | 2026-05-01T04:41:43Z | null | CONTRIBUTOR | null | 20260501T113057Z | 2026-05-01T11:30:57Z | kevinlin-openai | 177,054,202 | U_kgDOCo2h-g | User | false |
openclaw/openclaw | 4,164,957,808 | I_kwDOQb6kR874QDZw | 57,148 | https://github.com/openclaw/openclaw/issues/57148 | https://api.github.com/repos/openclaw/openclaw/issues/57148 | Feature: Add maxSubagentResultChars to truncate subagent results injected into parent session | ## Problem
When a subagent completes a task, its full output is injected into the parent session as a user message wrapped in `<<<BEGIN_UNTRUSTED_CHILD_RESULT>>>` markers. There is **no size limit** on this injection.
A single subagent analyzing an 849KB chat log returned 8.2K output tokens (~25K chars). In a session... | open | null | false | 2 | [] | [] | 2026-03-29T16:21:02Z | 2026-05-01T06:04:26Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | justin7974 | 46,263,464 | MDQ6VXNlcjQ2MjYzNDY0 | User | false |
openclaw/openclaw | 4,164,904,302 | I_kwDOQb6kR874P2Vu | 57,142 | https://github.com/openclaw/openclaw/issues/57142 | https://api.github.com/repos/openclaw/openclaw/issues/57142 | Webchat: autocomplete/suggestions for custom terms in chat input | ## Feature Request
**Problem:** When using dictation or typing in the webchat UI, project-specific terms (e.g. "MAVSAFE", "CAPCO", "Railway", "Supabase") are hard to input accurately. Dictation especially struggles with proper nouns and technical terms.
**Proposed Solution:** Add an autocomplete/suggestion system to ... | open | null | false | 2 | [
"stale"
] | [] | 2026-03-29T16:07:49Z | 2026-04-30T07:54:34Z | null | NONE | null | 20260430T090027Z | 2026-04-30T09:00:27Z | eliteweblabs | 1,345,837 | MDQ6VXNlcjEzNDU4Mzc= | User | false |
openclaw/openclaw | 4,163,348,829 | I_kwDOQb6kR874J6ld | 56,881 | https://github.com/openclaw/openclaw/issues/56881 | https://api.github.com/repos/openclaw/openclaw/issues/56881 | Regression after 2026.3.28: sessionStrategy behavior changed, ws-stream 500 fallback, slower Discord interaction handling | ## Summary
After upgrading OpenClaw from a working older build to **2026.3.28**, our local setup started behaving incorrectly. Rolling back to **2026.3.24** immediately restores normal behavior.
At the moment this looks like an OpenClaw regression rather than a problem in our local reranker server or in `memory-lance... | open | null | false | 3 | [] | [] | 2026-03-29T08:27:00Z | 2026-05-01T06:04:20Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jlin53882 | 76,562,059 | MDQ6VXNlcjc2NTYyMDU5 | User | false |
openclaw/openclaw | 4,164,225,965 | I_kwDOQb6kR874NQut | 57,053 | https://github.com/openclaw/openclaw/issues/57053 | https://api.github.com/repos/openclaw/openclaw/issues/57053 | feat(session): expose remaining session budget as structured data in session_status tool | ## Problem
When skills need to decide whether to spawn expensive sub-agent workflows (e.g., agent-debate spawns 6 sub-agents + judge = ~1400s total), there is no reliable programmatic way to check remaining session budget before committing to the spawn. The current `session_status` tool returns human-readable status t... | open | null | false | 2 | [] | [] | 2026-03-29T13:09:54Z | 2026-05-01T04:41:26Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | gasvn | 20,515,144 | MDQ6VXNlcjIwNTE1MTQ0 | User | false |
openclaw/openclaw | 4,163,939,474 | I_kwDOQb6kR874MKyS | 56,979 | https://github.com/openclaw/openclaw/issues/56979 | https://api.github.com/repos/openclaw/openclaw/issues/56979 | Auth provider token 过期时导致过多的模型回退日志 | ## 问题描述
当某个模型的认证 provider token 过期（例如 HTTP 401: invalid access token）时，系统会触发模型回退（fallback）逻辑。但这会导致每秒产生数十条 `model-fallback/decision` 日志，造成日志 spam。
## 影响
- 日志文件体积快速膨胀
- 难以查看其他有效日志
- 增加磁盘 I/O
## 日志示例
```
2026-03-29T19:24:34.304+08:00 [agent/embedded] embedded run agent end: ... error=HTTP 401: invalid access token or... | open | null | false | 2 | [
"stale"
] | [] | 2026-03-29T11:46:20Z | 2026-04-30T07:54:29Z | null | NONE | null | 20260430T090027Z | 2026-04-30T09:00:27Z | yanan1991 | 259,975,223 | U_kgDOD37oNw | User | false |
openclaw/openclaw | 4,163,166,436 | I_kwDOQb6kR874JODk | 56,856 | https://github.com/openclaw/openclaw/issues/56856 | https://api.github.com/repos/openclaw/openclaw/issues/56856 | feat(tui): clarify session state labels — distinguish Idle, Generating, Working, and Waiting states | ## Problem
The TUI currently shows **"Streaming | Connected"** for multiple distinct session states, making it impossible for operators to tell whether the agent is actually working or just idle with an open WebSocket.
Observed in practice: after a `sessions_yield` call (agent yielded waiting for a subagent to comple... | open | null | false | 2 | [
"stale"
] | [] | 2026-03-29T07:28:04Z | 2026-04-30T07:54:14Z | null | NONE | null | 20260430T090027Z | 2026-04-30T09:00:27Z | UGDev-Chane | 180,237,537 | U_kgDOCr404Q | User | false |
openclaw/openclaw | 4,163,313,118 | I_kwDOQb6kR874Jx3e | 56,876 | https://github.com/openclaw/openclaw/issues/56876 | https://api.github.com/repos/openclaw/openclaw/issues/56876 | [Feature Request] Add file type detection and preprocessing for binary documents | ## Problem Description
When reading binary files like .docx, .xlsx, .pdf through the read tool, the raw binary content is passed directly to the model as context. This causes several issues:
1. **Token Waste**: Binary data (e.g., "\x50\x4B\x03\x04...") consumes valuable context window tokens without providing meaning... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-29T08:15:40Z | 2026-05-03T04:38:02Z | 2026-04-30T08:01:43Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | ifitrain | 154,914,041 | U_kgDOCTvM-Q | User | false |
openclaw/openclaw | 4,163,924,282 | I_kwDOQb6kR874MHE6 | 56,973 | https://github.com/openclaw/openclaw/issues/56973 | https://api.github.com/repos/openclaw/openclaw/issues/56973 | refactor: remove normalizeChannelId wrapper collision — use canonical names directly | ## Summary
Two exported functions named `normalizeChannelId` with different return types force aliased imports at call sites that need both. The canonical, unambiguous names already exist and are widely used — the wrapper functions create unnecessary confusion.
## The collision
| Location | Returns | Delegates to |
... | open | null | false | 2 | [
"stale"
] | [] | 2026-03-29T11:42:30Z | 2026-04-30T07:54:26Z | null | NONE | null | 20260430T090027Z | 2026-04-30T09:00:27Z | jamie8johnson | 85,816,907 | MDQ6VXNlcjg1ODE2OTA3 | User | false |
openclaw/openclaw | 4,163,097,685 | I_kwDOQb6kR874I9RV | 56,850 | https://github.com/openclaw/openclaw/issues/56850 | https://api.github.com/repos/openclaw/openclaw/issues/56850 | Feature: Allow cron heartbeat interval to vary based on state conditions (deadline-aware scheduling) | ## Problem
OpenClaw cron jobs fire on fixed intervals (e.g., every 168h for weekly tasks). There is no mechanism to fire a cron job more frequently when a state condition is met — for example, when a prediction deadline is within 3 days, or when a pending review queue exceeds a threshold.
## Current Workaround
Agent... | open | null | false | 2 | [] | [] | 2026-03-29T07:04:53Z | 2026-05-01T06:04:19Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | gasvn | 20,515,144 | MDQ6VXNlcjIwNTE1MTQ0 | User | false |
openclaw/openclaw | 4,163,924,786 | I_kwDOQb6kR874MHMy | 56,974 | https://github.com/openclaw/openclaw/issues/56974 | https://api.github.com/repos/openclaw/openclaw/issues/56974 | refactor: consolidate duplicated fileExists implementations | ## Summary
Multiple files define their own local `fileExists` function when shared exports already exist. At least 2 are functionally identical to existing exports and can be replaced directly; 4 more use `fs.access` instead of `fs.stat` but are behaviorally equivalent for the read-only existence checks they perform.
... | open | null | false | 2 | [
"stale"
] | [] | 2026-03-29T11:42:39Z | 2026-04-30T07:54:28Z | null | NONE | null | 20260430T090027Z | 2026-04-30T09:00:27Z | jamie8johnson | 85,816,907 | MDQ6VXNlcjg1ODE2OTA3 | User | false |
openclaw/openclaw | 4,183,507,348 | I_kwDOQb6kR875W0GU | 58,660 | https://github.com/openclaw/openclaw/issues/58660 | https://api.github.com/repos/openclaw/openclaw/issues/58660 | openclawwechat 插件安装被误拦截 | 安装 openclawwechat 插件时，旧版本 openclaw-weixin 被误识别为需要更新，导致安装流程混乱。
**问题描述**：
1. 执行 Downloading openclawwechat…
Extracting /tmp/openclaw-npm-pack-K5orSR/openclawwechat-1.3.2.tgz…
Installing to /home/dpro/.openclaw/extensions/openclawwechat…
openclawwechat already at 1.3.2.
Restart the gateway to load plugins and hooks. 时，系统... | open | null | false | 3 | [] | [] | 2026-04-01T01:55:33Z | 2026-05-01T04:42:59Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | YangZH-hong | 215,106,874 | U_kgDODNJFOg | User | false |
openclaw/openclaw | 4,180,064,068 | I_kwDOQb6kR875JrdE | 58,433 | https://github.com/openclaw/openclaw/issues/58433 | https://api.github.com/repos/openclaw/openclaw/issues/58433 | [Bug]: [Windows] Path character loss bug - "system" becomes "ystem" in file paths | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
## Description
On Windows, when using file tools (read/write/edit), some characters in the path are being dropped/los
t. For example:
- Expected path: `D:\Code\knowledge-evolution-system\...`
- Actual path use... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T15:11:31Z | 2026-05-01T06:04:44Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | lfjwish | 31,098,841 | MDQ6VXNlcjMxMDk4ODQx | User | false |
openclaw/openclaw | 4,182,285,605 | I_kwDOQb6kR875SJ0l | 58,558 | https://github.com/openclaw/openclaw/issues/58558 | https://api.github.com/repos/openclaw/openclaw/issues/58558 | Support async handlers for tool_result_persist hook | ## Summary
The `tool_result_persist` hook currently has a synchronous-only signature:
```typescript
tool_result_persist: (event: PluginHookToolResultPersistEvent, ctx: PluginHookToolResultPersistContext)
=> PluginHookToolResultPersistResult | void;
```
This prevents plugins from making async operations (HTTP call... | open | null | false | 3 | [] | [] | 2026-03-31T20:53:25Z | 2026-05-01T06:04:52Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | saurabhjain1592 | 8,345,163 | MDQ6VXNlcjgzNDUxNjM= | User | false |
openclaw/openclaw | 4,175,792,512 | I_kwDOQb6kR8745YmA | 58,139 | https://github.com/openclaw/openclaw/issues/58139 | https://api.github.com/repos/openclaw/openclaw/issues/58139 | [Bug]: memory-lancedb plugin fails with Windows Docker bind mount | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
## Summary
The `memory-lancedb` plugin fails to initialize when its data directory is bind-mounted to a Windows host path. This is due to file system sync delays between Windows and Docker containers when usin... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T04:58:42Z | 2026-05-01T06:04:30Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | rudyshaw1988 | 246,858,920 | U_kgDODrbEqA | User | false |
openclaw/openclaw | 4,181,746,634 | I_kwDOQb6kR875QGPK | 58,537 | https://github.com/openclaw/openclaw/issues/58537 | https://api.github.com/repos/openclaw/openclaw/issues/58537 | [Bug]: Feishu oc_ (OpenChat) chats create duplicate group/direct sessions — outbound vs inbound chatType mismatch | # [Bug]: Feishu oc_ (OpenChat) chats create duplicate group/direct sessions — outbound vs inbound chatType mismatch
## Bug type
Bug
## Summary
When OpenClaw proactively sends messages to a Feishu OpenChat (oc_ prefix) via `message send` or cron delivery, the outbound path creates a **`direct`** session key (e.g. `a... | open | reopened | false | 2 | [] | [] | 2026-03-31T19:21:52Z | 2026-05-01T06:04:50Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | pc-seawind | 25,240,880 | MDQ6VXNlcjI1MjQwODgw | User | false |
openclaw/openclaw | 4,184,383,106 | I_kwDOQb6kR875aJ6C | 58,776 | https://github.com/openclaw/openclaw/issues/58776 | https://api.github.com/repos/openclaw/openclaw/issues/58776 | Sub-agent startup silence: no first-progress signal, aborted runs appear as hangs | ## Problem
Sub-agent runs can sit in `running` state indefinitely with zero output. From the parent's perspective this is indistinguishable from a healthy slow start. The actual cause is `openclaw:prompt-error` with `error: "aborted"` being mapped to `{ status: "timeout" }` — losing the distinction between startup abo... | open | null | false | 2 | [] | [] | 2026-04-01T06:07:18Z | 2026-05-01T04:43:14Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | SmithU64 | 117,943,562 | U_kgDOBwetCg | User | false |
openclaw/openclaw | 4,184,384,492 | I_kwDOQb6kR875aKPs | 58,777 | https://github.com/openclaw/openclaw/issues/58777 | https://api.github.com/repos/openclaw/openclaw/issues/58777 | Sub-agent model pin fidelity: requested vs effective model mismatch, no telemetry | ## Problem
When a sub-agent is spawned with an explicit model pin, the actual model that executes the run can differ silently. `effective_model`, `effective_provider`, and `effective_thinking` are never populated in routing telemetry or lifecycle events, so the parent cannot verify pin fidelity.
We identified **6 mis... | open | null | false | 2 | [] | [] | 2026-04-01T06:07:37Z | 2026-05-01T04:43:15Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | SmithU64 | 117,943,562 | U_kgDOBwetCg | User | false |
openclaw/openclaw | 4,178,488,779 | I_kwDOQb6kR875Dq3L | 58,351 | https://github.com/openclaw/openclaw/issues/58351 | https://api.github.com/repos/openclaw/openclaw/issues/58351 | workspaceOnly: true blocks agent from reading skill files | ## Problem
When an agent has `tools.fs.workspaceOnly: true`, the agent cannot read skill files because they are located outside the workspace at `~/.openclaw/skills/`.
The system prompt lists available skills with their **global paths** (e.g., `/home/user/.openclaw/skills/my-skill/SKILL.md`). When the agent tries to ... | open | null | false | 3 | [] | [] | 2026-03-31T11:48:50Z | 2026-05-01T06:04:39Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | sisutuulenisa | 261,256,991 | U_kgDOD5J3Hw | User | false |
openclaw/openclaw | 4,183,003,096 | I_kwDOQb6kR875U4_Y | 58,608 | https://github.com/openclaw/openclaw/issues/58608 | https://api.github.com/repos/openclaw/openclaw/issues/58608 | Session corruption from orphaned tool_result blocks survives restart (overload/timeout interruption) | ## Problem
When a session is interrupted mid-tool-call (overload, timeout, rate limit), the conversation history can end up with orphaned `tool_result` blocks that have no matching `tool_use` before them. On the next API call, Anthropic rejects the malformed history (e.g. `messages.36` validation error), and the sessi... | open | null | false | 3 | [] | [] | 2026-03-31T23:32:19Z | 2026-05-01T04:42:51Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | funkylazer | 164,404,385 | U_kgDOCcycoQ | User | false |
openclaw/openclaw | 4,175,143,662 | I_kwDOQb6kR87426Lu | 58,057 | https://github.com/openclaw/openclaw/issues/58057 | https://api.github.com/repos/openclaw/openclaw/issues/58057 | Feature: Dynamic identity resolution for allowlists (dmPolicy: dynamic) | ## Problem
The current `dmPolicy: "allowlist"` and `groupPolicy: "allowlist"` require hardcoded sender identifiers in `openclaw.json`. Every new user requires a config edit + gateway reload. This does not scale for multi-user deployments, especially on channels like BlueBubbles/iMessage where identifiers can be phone ... | open | null | false | 2 | [] | [] | 2026-03-31T02:01:17Z | 2026-05-01T04:41:52Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | mids-neo | 117,532,760 | U_kgDOBwFoWA | User | false |
openclaw/openclaw | 4,175,250,283 | I_kwDOQb6kR8743UNr | 58,067 | https://github.com/openclaw/openclaw/issues/58067 | https://api.github.com/repos/openclaw/openclaw/issues/58067 | [Bug]: Control UI frozen during sessions_yield + subagent wait — no progress indicator or streaming | ## Summary
When the main agent uses `sessions_spawn` + `sessions_yield` to delegate work to a subagent, **all connected surfaces** (Control UI webchat, TUI, etc.) appear completely frozen for the duration of `agent.wait`. No spinner, no progress indicator, no streaming chunks — the UI looks stuck/dead until the subage... | open | null | false | 2 | [] | [] | 2026-03-31T02:33:04Z | 2026-05-01T04:41:53Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | firstmovr | 155,568,852 | U_kgDOCUXK1A | User | false |
openclaw/openclaw | 4,175,927,506 | I_kwDOQb6kR87455jS | 58,147 | https://github.com/openclaw/openclaw/issues/58147 | https://api.github.com/repos/openclaw/openclaw/issues/58147 | Control UI / WebChat: Respect agents.defaults.timeFormat for session duration display | ## Description
The Control UI status bar displays session duration and timestamps without respecting the gents.defaults.timeFormat setting. Currently, duration is shown as 7:17 (12h-style, dropping leading zero) instead of | open | null | false | 2 | [] | [] | 2026-03-31T05:27:55Z | 2026-05-01T06:04:32Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | TommoT2 | 106,430,507 | U_kgDOBlgAKw | User | false |
openclaw/openclaw | 4,184,870,759 | I_kwDOQb6kR875cA9n | 58,832 | https://github.com/openclaw/openclaw/issues/58832 | https://api.github.com/repos/openclaw/openclaw/issues/58832 | [Slack] DM top-level messages incorrectly generate :thread: session keys with replyToMode != off | ## Bug Description
Top-level Slack **DM (direct message)** messages incorrectly receive a `:thread:` session key suffix when `replyToMode` is not `"off"`, identical to the channel bug fixed in #32353 / PR #32320.
The channel fix only addressed channel messages. DMs with `dmScope: "per-channel-peer"` still exhibit the... | closed | completed | true | 3 | [] | [] | 2026-04-01T07:41:48Z | 2026-05-04T04:33:55Z | 2026-05-02T04:17:35Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | daye-jjeong | 22,784,740 | MDQ6VXNlcjIyNzg0NzQw | User | false |
openclaw/openclaw | 4,181,242,334 | I_kwDOQb6kR875OLHe | 58,514 | https://github.com/openclaw/openclaw/issues/58514 | https://api.github.com/repos/openclaw/openclaw/issues/58514 | [Bug]: Google Chat: Space/Group messages silently ignored (DMs work correctly) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Google Chat Space/Group messages are received (HTTP 200) but silently ignored; DMs work correctly. Tested with both Add-on and standard Chat API apps.
### Steps to reproduce
1. Configure Google Chat webhook ... | open | null | false | 5 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T17:59:25Z | 2026-05-01T04:42:40Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | rubensandrade-sketch | 264,141,151 | U_kgDOD755Xw | User | false |
openclaw/openclaw | 4,181,045,395 | I_kwDOQb6kR875NbCT | 58,498 | https://github.com/openclaw/openclaw/issues/58498 | https://api.github.com/repos/openclaw/openclaw/issues/58498 | [Bug]: Status card shows the active OAuth profile, but weekly usage appears to reflect a different Oauth credential | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The status card shows the correct active OAuth profile for ChatGPT Codex, but the weekly usage line reports 0% left even though that limit appears to belong to a different credential.
### Steps to reproduce
1... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T17:34:06Z | 2026-05-01T06:04:47Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | QuantumScritches | 264,891,166 | U_kgDOD8nrHg | User | false |
openclaw/openclaw | 4,179,615,949 | I_kwDOQb6kR875H-DN | 58,407 | https://github.com/openclaw/openclaw/issues/58407 | https://api.github.com/repos/openclaw/openclaw/issues/58407 | [Feature]: Inject parent channel pinned messages into Discord thread session context | ## Feature Request
### Problem
When a new thread is created in a Discord server channel, the agent spawns a fresh session with no awareness of the parent channel's pinned messages. Many users (ourselves included) use pinned messages to store channel-specific operating rules, workflow instructions, and project context ... | open | null | false | 2 | [] | [] | 2026-03-31T14:17:41Z | 2026-05-01T06:04:42Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | garyz-inspirum | 174,663,437 | U_kgDOCmknDQ | User | false |
openclaw/openclaw | 4,183,761,352 | I_kwDOQb6kR875XyHI | 58,702 | https://github.com/openclaw/openclaw/issues/58702 | https://api.github.com/repos/openclaw/openclaw/issues/58702 | [Bug] WebChat: Message text covered by action icons after upgrade to 2026.3.31 | ## Bug Description
After upgrading to 2026.3.31, the WebChat UI shows a layout issue where the action icons (copy/expand buttons) that should only appear on hover are permanently displayed and covering the end of message text.
## Steps to Reproduce
1. Open WebChat interface (http://127.0.0.1:18789)
2. Send any message... | open | null | false | 4 | [] | [] | 2026-04-01T03:19:15Z | 2026-05-01T04:43:04Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | warpig2023 | 124,851,041 | U_kgDOB3ETYQ | User | false |
openclaw/openclaw | 4,183,915,216 | I_kwDOQb6kR875YXrQ | 58,730 | https://github.com/openclaw/openclaw/issues/58730 | https://api.github.com/repos/openclaw/openclaw/issues/58730 | [Feature Request] exec() sandbox isolation and tool permission model — lessons from Claude Code source leak | ## Motivation
After analyzing the leaked Claude Code source (v2.1.88, March 2026), several architectural patterns stand out as highly relevant to improving OpenClaw. This issue consolidates actionable feature requests inspired by that analysis.
---
## 1. exec() Sandbox Isolation (High Priority — Security)
**Problem... | open | null | false | 3 | [] | [] | 2026-04-01T04:08:31Z | 2026-05-01T04:43:08Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | Bobby-Bot-00 | 261,764,971 | U_kgDOD5o3aw | User | false |
openclaw/openclaw | 4,175,237,199 | I_kwDOQb6kR8743RBP | 58,063 | https://github.com/openclaw/openclaw/issues/58063 | https://api.github.com/repos/openclaw/openclaw/issues/58063 | web_search tool aborts immediately with xAI Responses API (2026.3.28) | ## Bug Description
The `web_search` tool aborts immediately (no timeout, no error log) when configured with `tools.web.search.provider: grok` after updating to 2026.3.28.
**Expected behavior:** Grok web search returns AI-synthesized search results.
**Actual behavior:** Tool returns `This operation was aborted` within ... | closed | completed | true | 4 | [] | [] | 2026-03-31T02:28:29Z | 2026-05-04T04:34:40Z | 2026-05-02T01:56:01Z | NONE | null | 20260504T060031Z | 2026-05-04T06:00:31Z | marvcasasola-svg | 267,465,534 | U_kgDOD_EzPg | User | false |
openclaw/openclaw | 4,175,432,185 | I_kwDOQb6kR8744An5 | 58,106 | https://github.com/openclaw/openclaw/issues/58106 | https://api.github.com/repos/openclaw/openclaw/issues/58106 | OpenRouter models hidden in model picker — PI auth discovery fails to resolve keyRef auth profiles | ## Problem
OpenRouter models are configured correctly in `openclaw.json` and appear in per-agent `models.json` files, but they do not appear in the Discord `/models` picker or text `/model` command. The picker shows the full OpenRouter catalog (via direct API discovery) but treats all models as unauthenticated — so se... | open | null | false | 3 | [] | [] | 2026-03-31T03:25:48Z | 2026-05-01T04:41:57Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | ThalynLabs | 38,703,365 | MDQ6VXNlcjM4NzAzMzY1 | User | false |
openclaw/openclaw | 4,183,814,955 | I_kwDOQb6kR875X_Mr | 58,713 | https://github.com/openclaw/openclaw/issues/58713 | https://api.github.com/repos/openclaw/openclaw/issues/58713 | Feature Request: Ephemeral sessions for agent-to-agent communication | ## Problem
When agent A sends a message to agent B via `sessions_send`, the message is injected into B's existing main session. This pollutes B's conversation context — mixing user-facing chat history with inter-agent coordination messages.
The same issue happens in reverse: when B replies to A, the reply lands in A'... | open | null | false | 2 | [] | [] | 2026-04-01T03:37:30Z | 2026-05-01T04:43:06Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | hwj123hwj | 130,386,673 | U_kgDOB8WK8Q | User | false |
openclaw/openclaw | 4,183,941,224 | I_kwDOQb6kR875YeBo | 58,737 | https://github.com/openclaw/openclaw/issues/58737 | https://api.github.com/repos/openclaw/openclaw/issues/58737 | [Bug]: Agent display name/avatar reverts to bot default on edited Slack messages | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
When an agent posts a message to Slack with a custom ```username``` and ```icon_url``` override (e.g. Name in #agent-name), the display name and avatar show correctly on the initial post. However, when OpenClaw edits that me... | open | null | false | 4 | [
"bug",
"regression"
] | [] | 2026-04-01T04:17:04Z | 2026-05-01T04:43:12Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | FergusClare | 19,366,735 | MDQ6VXNlcjE5MzY2NzM1 | User | false |
openclaw/openclaw | 4,185,336,537 | I_kwDOQb6kR875dyrZ | 58,890 | https://github.com/openclaw/openclaw/issues/58890 | https://api.github.com/repos/openclaw/openclaw/issues/58890 | [Bug] Auto-update subprocess crashes before restart handoff: piped stdout/stderr breaks when bootout kills parent gateway | ## Summary
The built-in auto-update consistently kills the gateway on macOS with no recovery, leaving the service unloaded indefinitely. The root cause is that `resolveCommandStdio` forces piped stdout/stderr for the update subprocess, creating a fatal dependency on the gateway parent process — which is then killed by... | open | null | false | 5 | [] | [] | 2026-04-01T09:02:30Z | 2026-05-01T04:43:26Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | moxy-kit | 261,167,793 | U_kgDOD5EasQ | User | false |
openclaw/openclaw | 4,183,930,735 | I_kwDOQb6kR875Ybdv | 58,733 | https://github.com/openclaw/openclaw/issues/58733 | https://api.github.com/repos/openclaw/openclaw/issues/58733 | [Bug]: bundled Grok web_search inherits a too-short 30s default timeout on the xAI Responses path | ## Bug type
Runtime / provider timeout default bug
## Summary
The bundled Grok `web_search` path on current `upstream/main` still inherits the generic shared web tool default timeout of 30 seconds.
In local verification against the bundled xAI Responses path, that default is too short for at least some Grok web sea... | closed | completed | true | 3 | [] | [] | 2026-04-01T04:13:36Z | 2026-05-04T04:34:38Z | 2026-05-02T01:56:09Z | CONTRIBUTOR | null | 20260504T060031Z | 2026-05-04T06:00:31Z | Nanako0129 | 44,753,291 | MDQ6VXNlcjQ0NzUzMjkx | User | false |
openclaw/openclaw | 4,177,691,778 | I_kwDOQb6kR875AoSC | 58,284 | https://github.com/openclaw/openclaw/issues/58284 | https://api.github.com/repos/openclaw/openclaw/issues/58284 | Feature request: turn-based / rolling session pruning (not only cache-ttl) | ## Problem
Session pruning currently only runs when the session has been **inactive longer than the TTL** (`contextPruning.ttl`). In active sessions where messages flow continuously (every 30s-1min), the TTL never expires and pruning **never triggers**.
This means tool results accumulate indefinitely in the context d... | open | null | false | 2 | [] | [] | 2026-03-31T09:46:57Z | 2026-05-01T06:04:36Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | rtmuller | 66,280,176 | MDQ6VXNlcjY2MjgwMTc2 | User | false |
openclaw/openclaw | 4,177,764,002 | I_kwDOQb6kR875A56i | 58,289 | https://github.com/openclaw/openclaw/issues/58289 | https://api.github.com/repos/openclaw/openclaw/issues/58289 | [Bug]: Non-main agent may miss models.json and fail with "Unknown model" when provider plan is empty | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When a non-main agent has no models.json and provider resolution returns empty, startup skips models.json generation and later fails with "Unknown model".
### Steps to reproduce
1. Prepare main agent with a v... | open | null | false | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T09:58:27Z | 2026-05-01T06:04:37Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | xiangri678 | 90,504,308 | MDQ6VXNlcjkwNTA0MzA4 | User | false |
openclaw/openclaw | 4,182,514,527 | I_kwDOQb6kR875TBtf | 58,570 | https://github.com/openclaw/openclaw/issues/58570 | https://api.github.com/repos/openclaw/openclaw/issues/58570 | Gateway should log warning when message is dropped due to allow: false | ## Summary
When a message resolves to a channel config entry that has `allow: false`, the gateway drops the message silently — no log output, no warning, nothing. This makes silent policy denials indistinguishable from transport or routing bugs during incident response.
## Requested behavior
When a message is droppe... | open | null | false | 2 | [] | [] | 2026-03-31T21:33:42Z | 2026-05-01T06:04:53Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | UBIpromoter | 46,543,177 | MDQ6VXNlcjQ2NTQzMTc3 | User | false |
openclaw/openclaw | 4,175,356,638 | I_kwDOQb6kR8743uLe | 58,088 | https://github.com/openclaw/openclaw/issues/58088 | https://api.github.com/repos/openclaw/openclaw/issues/58088 | enforceSessionDiskBudget (and other maintenance paths) evict topic/thread session pointers | ## Summary
`enforceSessionDiskBudget`, `pruneStaleEntries`, and `capEntryCount` in `auth-profiles-*.js` can all delete topic and thread session entries from `sessions.json`. These are long-lived group conversation sessions that users expect to persist indefinitely. Losing them causes permanent context loss with no war... | closed | completed | true | 3 | [] | [] | 2026-03-31T03:05:08Z | 2026-05-05T04:28:40Z | 2026-05-02T05:31:09Z | NONE | null | 20260505T060034Z | 2026-05-05T06:00:34Z | drinkflav | 256,454,846 | U_kgDOD0kwvg | User | false |
openclaw/openclaw | 4,170,772,851 | I_kwDOQb6kR874mPFz | 57,713 | https://github.com/openclaw/openclaw/issues/57713 | https://api.github.com/repos/openclaw/openclaw/issues/57713 | [Bug]: Default sandbox image openclaw-sandbox:bookworm-slim lacked python3, breaking edit/write | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
Version
• OpenClaw updated to 2026.3.28
Symptoms
• read worked
• edit failed
• write failed
Observed errors
Gateway log showed:
• error [tools] edit failed: moltbot-sandbox-fs: 2: python3: not found
• error [tools] writ... | open | null | false | 2 | [
"bug"
] | [] | 2026-03-30T13:31:03Z | 2026-05-01T04:41:47Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | PraxoOnline | 257,132,519 | U_kgDOD1OH5w | User | false |
openclaw/openclaw | 4,185,319,738 | I_kwDOQb6kR875duk6 | 58,887 | https://github.com/openclaw/openclaw/issues/58887 | https://api.github.com/repos/openclaw/openclaw/issues/58887 | feat(typing): show typing indicator immediately on voice message receipt, before STT transcription | ## Problem
When a user sends a voice message, the typing indicator only appears **after** STT transcription completes — not when the message is received. This creates a silent wait of 3–6 seconds before any feedback is shown:
| Step | Time |
|------|------|
| Telegram polling + message received | ~1s |
| Audio file d... | open | null | false | 3 | [] | [] | 2026-04-01T09:00:03Z | 2026-05-01T04:43:24Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | richardhxwang | 225,731,730 | U_kgDODXRkkg | User | false |
openclaw/openclaw | 4,177,568,962 | I_kwDOQb6kR875AKTC | 58,272 | https://github.com/openclaw/openclaw/issues/58272 | https://api.github.com/repos/openclaw/openclaw/issues/58272 | [Bug]: macOS Path Resolution: ENOENT on both mixed-character directories AND mixed-character filenames (CJK + ASCII) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
On macOS, absolute path resolution fails completely when any directory in the path OR the final filename itself contains a mixture of CJK (Chinese) and ASCII (English/numeric) characters. Both scenarios return ... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-31T09:28:49Z | 2026-05-01T06:04:34Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | yutrinapo-cmyk | 272,461,481 | U_kgDOED1uqQ | User | false |
openclaw/openclaw | 4,179,755,556 | I_kwDOQb6kR875IgIk | 58,414 | https://github.com/openclaw/openclaw/issues/58414 | https://api.github.com/repos/openclaw/openclaw/issues/58414 | [Bug] Gateway crash loop after self-update: stdin hang + v2026.3.28 `gateway` command exits 0 without starting server | ## Summary
After updating OpenClaw via the agent itself (`npm i -g openclaw@beta`), the gateway entered an unrecoverable crash loop lasting **16+ hours**. Two independent bugs combined to prevent any recovery — including by the built-in watchdog.
**Affected versions:** 2026.3.24, 2026.3.28 (current `latest` and `beta... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-31T14:36:29Z | 2026-05-03T04:38:01Z | 2026-04-30T08:01:51Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | arcabotai | 261,107,832 | U_kgDOD5AweA | User | false |
openclaw/openclaw | 4,179,695,163 | I_kwDOQb6kR875IRY7 | 58,411 | https://github.com/openclaw/openclaw/issues/58411 | https://api.github.com/repos/openclaw/openclaw/issues/58411 | sessions_spawn lacks --bind here equivalent — agent cannot bind ACP session to existing Discord thread | ## Summary
When a user is in an existing Discord thread and asks the agent to start a coding session, the correct UX is `/acp spawn <agent> --bind here` — which attaches the ACP session to the current thread in place.
However, this slash command can only be run by a human. If the user asks the **agent** to spawn the ... | open | null | false | 2 | [] | [] | 2026-03-31T14:28:23Z | 2026-05-01T06:04:43Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | brew-agent | 263,394,628 | U_kgDOD7MVRA | User | false |
openclaw/openclaw | 4,184,303,022 | I_kwDOQb6kR875Z2Wu | 58,765 | https://github.com/openclaw/openclaw/issues/58765 | https://api.github.com/repos/openclaw/openclaw/issues/58765 | Feature: Support output dimensionality truncation for local GGUF embedding models | ## Summary
The `outputDimensionality` config field currently only works for Gemini embedding-2 models. Local GGUF embedding models (e.g., Qwen3-Embedding-0.6B) always output their full dimension (1024), with no way to truncate.
## Use Case
Many modern embedding models support Matryoshka Representation Learning (MRL)... | open | null | false | 2 | [] | [] | 2026-04-01T05:47:06Z | 2026-05-01T04:43:13Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | losz5000 | 24,774,426 | MDQ6VXNlcjI0Nzc0NDI2 | User | false |
openclaw/openclaw | 4,163,348,017 | I_kwDOQb6kR874J6Yx | 56,880 | https://github.com/openclaw/openclaw/issues/56880 | https://api.github.com/repos/openclaw/openclaw/issues/56880 | Concurrent message handling per session (async agent turns) | ### Problem
Currently, when an agent is processing a message on Discord, all subsequent messages from the same user/channel are queued and wait until the current turn completes before processing begins. This creates a significant UX bottleneck:
**Scenario:**
1. User asks a complex question (e.g., debugging a config i... | open | null | false | 2 | [
"stale"
] | [] | 2026-03-29T08:26:42Z | 2026-04-30T07:54:22Z | null | NONE | null | 20260430T090027Z | 2026-04-30T09:00:27Z | darealgege | 30,112,470 | MDQ6VXNlcjMwMTEyNDcw | User | false |
openclaw/openclaw | 4,355,482,087 | I_kwDOQb6kR88AAAABA5th5w | 74,869 | https://github.com/openclaw/openclaw/issues/74869 | https://api.github.com/repos/openclaw/openclaw/issues/74869 | Request: iOS Node TestFlight invite | Hi OpenClaw team,
Requesting TestFlight access for the official iOS Node app.
**Apple ID email (for TestFlight invite):** 3477492305@qq.com
**Current setup:**
Running OpenClaw Gateway on WSL2 (Ubuntu) with an Intel Core Ultra 9 285H + 32GB RAM. Infrastructure includes a homelab NAS with Cloudflare Zero Trust tunnels... | closed | not_planned | true | 1 | [] | [] | 2026-04-30T05:17:43Z | 2026-05-03T04:38:11Z | 2026-04-30T06:43:35Z | NONE | null | 20260503T060030Z | 2026-05-03T06:00:30Z | programmingWTF | 114,961,940 | U_kgDOBtouFA | User | false |
openclaw/openclaw | 4,183,904,923 | I_kwDOQb6kR875YVKb | 58,726 | https://github.com/openclaw/openclaw/issues/58726 | https://api.github.com/repos/openclaw/openclaw/issues/58726 | Loop detector should hard-abort agent after N identical tool calls, not just warn | ## Problem
The loop detector currently warns at 10/20/30 identical tool calls but **never force-stops the agent**. Tonight a session called `session_status` 30+ times in a row. The detector printed warnings at every milestone but the agent kept going, creating a stuck session for 12+ minutes.
## Current behavior
- Lo... | open | null | false | 2 | [] | [] | 2026-04-01T04:05:20Z | 2026-05-01T04:43:07Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dexbot17 | 259,017,668 | U_kgDOD3BLxA | User | false |
openclaw/openclaw | 4,182,634,469 | I_kwDOQb6kR875Te_l | 58,580 | https://github.com/openclaw/openclaw/issues/58580 | https://api.github.com/repos/openclaw/openclaw/issues/58580 | Session health check: auto-clear stuck sessions after prolonged unresponsiveness | ## Problem
After gateway crashes and restarts, sessions can accumulate corrupted state — pinned to a broken model override, bloated token counts (280k+), or stale context. The session appears \alive\ but silently fails on every incoming message.
In a real incident, 20+ Telegram messages went unanswered over several h... | open | null | false | 3 | [] | [] | 2026-03-31T22:00:08Z | 2026-05-01T06:04:55Z | null | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | jpeghead | 7,033,691 | MDQ6VXNlcjcwMzM2OTE= | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.