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 | 4,033,927,012 | I_kwDOQb6kR87wcNdk | 37,765 | https://github.com/openclaw/openclaw/issues/37765 | https://api.github.com/repos/openclaw/openclaw/issues/37765 | Document previous_response_id behavior in OpenResponses API — silently ignored | ## Summary
When using the `/v1/responses` endpoint, the `previous_response_id` parameter (as specified in the OpenAI Responses API) is silently accepted but does not provide conversation continuity. The `user` field is the correct mechanism for session management in OpenClaw, but this is not documented.
## Steps to R... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T12:20:44Z | 2026-04-27T05:02:17Z | 2026-04-24T19:28:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | asenden | 4,030,809 | MDQ6VXNlcjQwMzA4MDk= | User | false |
openclaw/openclaw | 4,033,956,790 | I_kwDOQb6kR87wcUu2 | 37,804 | https://github.com/openclaw/openclaw/issues/37804 | https://api.github.com/repos/openclaw/openclaw/issues/37804 | RFC: treat models.providers-injected models as first-class forward-compat models | ### Summary
OpenClaw already has the core pieces for forward-compat model support:
- `models.providers.<provider>.models` can inject model definitions from config
- runtime can resolve some newer models via forward-compat fallbacks
- `models status` can often resolve models that are not yet present in the underlying ... | closed | completed | false | 5 | [] | [] | 2026-03-06T12:26:51Z | 2026-05-04T10:19:29Z | 2026-05-04T10:19:29Z | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Sapientropic | 194,952,804 | U_kgDOC56-ZA | User | false |
openclaw/openclaw | 4,034,013,975 | I_kwDOQb6kR87wcisX | 37,813 | https://github.com/openclaw/openclaw/issues/37813 | https://api.github.com/repos/openclaw/openclaw/issues/37813 | [Bug]: Unrecognised model IDs silently fall back to primary default — bypasses configured fallback chain and tool permissions | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Model overrides for IDs not in the internal registry silently fall back to the primary default with full permissions, bypassing both the configured fallback chain and agent-visible status reporting.
### Steps to reproduce
1. Configure a ... | closed | completed | true | 4 | [
"bug",
"bug:behavior"
] | [] | 2026-03-06T12:37:55Z | 2026-05-01T04:40:27Z | 2026-04-29T03:50:44Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | Workz777 | 68,589,694 | MDQ6VXNlcjY4NTg5Njk0 | User | false |
openclaw/openclaw | 4,034,031,458 | I_kwDOQb6kR87wcm9i | 37,815 | https://github.com/openclaw/openclaw/issues/37815 | https://api.github.com/repos/openclaw/openclaw/issues/37815 | Feature Request: Native prompt scrubbing / PII redaction before sending to LLM providers | ## Summary
Add a native, configurable find-and-replace mechanism that scrubs sensitive terms from prompts **before** they are sent to LLM providers, and re-substitutes them in responses.
## Problem
OpenClaw agents with access to corporate data (emails, calendars, team chats, CRM) inevitably include company names, em... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-06T12:42:07Z | 2026-04-28T04:43:43Z | 2026-04-26T01:32:44Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | ImWally26 | 260,066,078 | U_kgDOD4BLHg | User | false |
openclaw/openclaw | 4,034,060,682 | I_kwDOQb6kR87wcuGK | 37,816 | https://github.com/openclaw/openclaw/issues/37816 | https://api.github.com/repos/openclaw/openclaw/issues/37816 | [Feature]: Telegram — full agent activity visibility (Web UI parity) | ## Feature Request
### Problem
The Web UI (Control Panel) shows everything the agent does in real time: thinking steps, tool calls, intermediate results, reasoning, memory operations. On Telegram, the user only sees "Nova is typing..." and then the final answer. All the intermediate activity is invisible.
This creat... | open | null | false | 2 | [] | [] | 2026-03-06T12:48:08Z | 2026-05-04T09:03:03Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | DasOif | 181,653,786 | U_kgDOCtPRGg | User | false |
openclaw/openclaw | 4,034,150,968 | I_kwDOQb6kR87wdEI4 | 37,832 | https://github.com/openclaw/openclaw/issues/37832 | https://api.github.com/repos/openclaw/openclaw/issues/37832 | WhatsApp/web-outbound: message tool media from workspace-<agentId> rejected (path-not-allowed) | ## Summary
When running an untrusted agent in Docker sandbox mode, media files generated in the agent workspace are blocked on WhatsApp outbound send with:
`LocalMediaAccessError: Local media path is not under an allowed directory`
Main error code is `path-not-allowed`.
## Context
Goal is to keep `tools.exec` sand... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T13:09:29Z | 2026-04-27T05:02:16Z | 2026-04-24T19:28:53Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | heyAyushh | 17,379,963 | MDQ6VXNlcjE3Mzc5OTYz | User | false |
openclaw/openclaw | 4,034,153,969 | I_kwDOQb6kR87wdE3x | 37,833 | https://github.com/openclaw/openclaw/issues/37833 | https://api.github.com/repos/openclaw/openclaw/issues/37833 | Feature Request: `agent:turn:end` internal hook event | ## Summary
Add an `agent:turn:end` internal hook event that fires once after a complete agent turn cycle (user message → agent processing → all tool calls → delivery) has finished.
## Motivation
The gateway already tracks turn completion internally via `AcpSessionManager.recordTurnCompletion()`, including duration, ... | closed | not_planned | true | 4 | [
"stale"
] | [] | 2026-03-06T13:10:12Z | 2026-04-28T04:43:42Z | 2026-04-26T01:32:53Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | mantisai-bot | 259,690,819 | U_kgDOD3qRQw | User | false |
openclaw/openclaw | 4,034,157,973 | I_kwDOQb6kR87wdF2V | 37,834 | https://github.com/openclaw/openclaw/issues/37834 | https://api.github.com/repos/openclaw/openclaw/issues/37834 | Session context corruption: orphaned tool_use ID causes permanent 400 loop after abort | ## Bug Report
**OpenClaw Version:** 2026.3.2 (inferred from recent upgrade)
**Session ID:** `2d882ee3-8503-4685-8eff-1cffc1c5835c`
**Provider/Model:** `anthropic/claude-sonnet-4-6`
**Date:** 2026-03-06 ~12:28 EST
---
## Summary
A long-running EPIC-002 Hive build session entered a permanent, unrecoverable error loop... | closed | completed | true | 12 | [
"stale"
] | [] | 2026-03-06T13:11:12Z | 2026-04-28T04:43:41Z | 2026-04-26T01:33:00Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | bryanbaer | 22,436,505 | MDQ6VXNlcjIyNDM2NTA1 | User | false |
openclaw/openclaw | 4,034,220,204 | I_kwDOQb6kR87wdVCs | 37,841 | https://github.com/openclaw/openclaw/issues/37841 | https://api.github.com/repos/openclaw/openclaw/issues/37841 | [Bug]: Session text invisible to user when message tool sends media on WhatsApp | ## Bug type
Behavior bug (silent data loss)
## Summary
When an agent writes session text and also calls the `message` tool to send media in the same turn, the session text is never delivered to the user on WhatsApp. No error is raised. The agent appears to have responded but the user receives only the media.
## Ste... | closed | not_planned | true | 3 | [] | [] | 2026-03-06T13:24:43Z | 2026-05-01T04:41:04Z | 2026-04-29T00:08:05Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | MaxHermez | 29,142,449 | MDQ6VXNlcjI5MTQyNDQ5 | User | false |
openclaw/openclaw | 4,034,220,787 | I_kwDOQb6kR87wdVLz | 37,842 | https://github.com/openclaw/openclaw/issues/37842 | https://api.github.com/repos/openclaw/openclaw/issues/37842 | Feature: Graph-aware loop detection for sessions_send (triangular/polygonal cascades bypass maxPingPongTurns) | ## Problem
`maxPingPongTurns` only tracks **direct pairwise** message exchanges between two agents (A<>B). It does not detect **triangular or polygonal** loops where messages cycle through 3+ agents (e.g., A->B->C->A). Each leg of a triangle appears as "turn 1" from that pair's perspective, so the guard never fires.
... | open | null | false | 3 | [] | [] | 2026-03-06T13:24:50Z | 2026-05-04T06:15:04Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | sprfrkr | 5,494,966 | MDQ6VXNlcjU0OTQ5NjY= | User | false |
openclaw/openclaw | 4,034,221,859 | I_kwDOQb6kR87wdVcj | 37,843 | https://github.com/openclaw/openclaw/issues/37843 | https://api.github.com/repos/openclaw/openclaw/issues/37843 | Cron Job Model Should Support default as Special Value for Runtime Resolution | ## Problem Description
Currently, when creating a cron job with agentTurn payload, the model field is saved as-is at creation time. This causes several issues:
1. Token expiration: If a user specifies --model volcengine/ark-code-latest and the token later expires, all cron jobs using that model fail
2. Provider chang... | open | null | false | 3 | [] | [] | 2026-03-06T13:25:02Z | 2026-05-04T06:15:05Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | ramsaycute | 201,072,219 | U_kgDOC_weWw | User | false |
openclaw/openclaw | 4,034,236,065 | I_kwDOQb6kR87wdY6h | 37,846 | https://github.com/openclaw/openclaw/issues/37846 | https://api.github.com/repos/openclaw/openclaw/issues/37846 | EROFS: Isolated cron session sandbox tries to write /file.lock at root (read-only), crashes before script executes | ## Bug Report
**Symptom:** Cron jobs using `sessionTarget: isolated` fail immediately with:
```
Error: EROFS: read-only file system, open '/file.lock'
```
**Timing:** Fails in ~148ms -- before any user script executes. This is a sandbox initialization error, not a script error.
**Root cause:** The isolated session s... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T13:27:59Z | 2026-04-27T05:02:14Z | 2026-04-24T19:29:01Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mmartoccia | 4,631,577 | MDQ6VXNlcjQ2MzE1Nzc= | User | false |
openclaw/openclaw | 4,034,258,194 | I_kwDOQb6kR87wdeUS | 37,852 | https://github.com/openclaw/openclaw/issues/37852 | https://api.github.com/repos/openclaw/openclaw/issues/37852 | [Bug] Synology Chat plugin: sendText/sendMedia fails with 'incoming URL not configured' due to empty config fallback | ## Bug Description
The `sendText` and `sendMedia` functions in the Synology Chat plugin (`extensions/synology-chat/src/channel.ts`) fail when called by the delivery recovery system because they fall back to `resolveAccount({}, accountId)` with an empty config object `{}`.
This causes `incomingUrl` to resolve as an em... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T13:33:09Z | 2026-04-27T05:02:14Z | 2026-04-24T19:29:06Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | vecsop | 31,853,196 | MDQ6VXNlcjMxODUzMTk2 | User | false |
openclaw/openclaw | 4,034,292,401 | I_kwDOQb6kR87wdmqx | 37,857 | https://github.com/openclaw/openclaw/issues/37857 | https://api.github.com/repos/openclaw/openclaw/issues/37857 | feishu-plugin-onboard doctor --fix disables most stock plugins (OpenClaw 2026.3.2) | ## Summary
Running `feishu-plugin-onboard doctor --fix` on OpenClaw `2026.3.2` unexpectedly disables almost all stock plugins/channels. After the fix, only `feishu-openclaw-plugin` remains loaded.
This appears to mutate plugin enablement state globally, not just `plugins.allow` for the Feishu plugin.
## Environment
-... | closed | completed | true | 2 | [
"stale"
] | [] | 2026-03-06T13:41:22Z | 2026-04-27T05:02:13Z | 2026-04-24T19:29:11Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lamkan0210 | 65,824,253 | MDQ6VXNlcjY1ODI0MjUz | User | false |
openclaw/openclaw | 4,034,315,549 | I_kwDOQb6kR87wdsUd | 37,865 | https://github.com/openclaw/openclaw/issues/37865 | https://api.github.com/repos/openclaw/openclaw/issues/37865 | [Bug]: TUI and web UI: streaming text does not actually stream when using tools, it only shows up when finalized | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
Previous issue of stream deletions https://github.com/openclaw/openclaw/issues/28180 was closed by https://github.com/openclaw/openclaw/commit/d326861eb470646ea1d419f733406b7bb75e5ebe
Technically it fixes the streaming deletions, but it do... | closed | completed | true | 3 | [
"bug",
"stale",
"bug:behavior"
] | [] | 2026-03-06T13:46:58Z | 2026-04-27T05:02:12Z | 2026-04-24T19:29:16Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | dalehamel | 618,615 | MDQ6VXNlcjYxODYxNQ== | User | false |
openclaw/openclaw | 4,034,341,172 | I_kwDOQb6kR87wdyk0 | 37,874 | https://github.com/openclaw/openclaw/issues/37874 | https://api.github.com/repos/openclaw/openclaw/issues/37874 | TUI incorrectly tags messages with channel: telegram causing cross-delivery | ## Bug Description
When using the TUI (`openclaw tui`), messages are incorrectly tagged with `channel: telegram` in the inbound metadata, even though `provider: webchat` and `surface: webchat` are correct.
This causes replies to be delivered to **both** the TUI (inline) and Telegram, resulting in duplicate messages.
... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T13:53:12Z | 2026-04-27T05:02:11Z | 2026-04-24T19:29:25Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | 12th-man | 39,319,810 | MDQ6VXNlcjM5MzE5ODEw | User | false |
openclaw/openclaw | 4,034,357,382 | I_kwDOQb6kR87wd2iG | 37,878 | https://github.com/openclaw/openclaw/issues/37878 | https://api.github.com/repos/openclaw/openclaw/issues/37878 | model-usage: handle numeric-string costs and ignore non-finite values | ## Summary
The `skills/model-usage/scripts/model_usage.py` helper currently aggregates costs only when values are native numbers.
In practice, usage payloads may include numeric cost fields serialized as strings (for example `"1.75"`). Those valid values can be dropped during aggregation/model selection. The script ca... | open | null | false | 2 | [] | [] | 2026-03-06T13:57:01Z | 2026-05-04T06:15:07Z | null | CONTRIBUTOR | null | 20260504T090105Z | 2026-05-04T09:01:05Z | shuofengzhang | 24,763,026 | MDQ6VXNlcjI0NzYzMDI2 | User | false |
openclaw/openclaw | 4,034,478,457 | I_kwDOQb6kR87weUF5 | 37,902 | https://github.com/openclaw/openclaw/issues/37902 | https://api.github.com/repos/openclaw/openclaw/issues/37902 | sessions_spawn: model overrides runTimeoutSeconds with decreasing values across retries; no config minimum enforced | ## Summary
When an agent calls `sessions_spawn` to delegate to a subagent, the calling model autonomously sets `runTimeoutSeconds` in the tool call. This overrides the `agents.defaults.subagents.runTimeoutSeconds` config value, which only acts as a **fallback default** — not a floor/minimum. The result is subagents re... | open | null | false | 3 | [] | [] | 2026-03-06T14:21:56Z | 2026-05-04T06:15:08Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | unw1red | 45,374,291 | MDQ6VXNlcjQ1Mzc0Mjkx | User | false |
openclaw/openclaw | 4,034,515,946 | I_kwDOQb6kR87wedPq | 37,907 | https://github.com/openclaw/openclaw/issues/37907 | https://api.github.com/repos/openclaw/openclaw/issues/37907 | Mutating tool warning fires even when agent successfully recovered in the same turn | ## Problem
When a mutating tool (e.g. `edit`) fails and the agent immediately retries with a corrected call that succeeds — all within the same turn — OpenClaw still posts the warning to the chat surface:
> ⚠️ 📝 Edit: `<path>` failed
The user sees a failure notification even though the action completed successfully... | closed | completed | true | 4 | [
"stale"
] | [] | 2026-03-06T14:29:00Z | 2026-04-27T05:02:10Z | 2026-04-24T19:29:30Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | haynzz | 1,236,319 | MDQ6VXNlcjEyMzYzMTk= | User | false |
openclaw/openclaw | 4,034,531,806 | I_kwDOQb6kR87wehHe | 37,913 | https://github.com/openclaw/openclaw/issues/37913 | https://api.github.com/repos/openclaw/openclaw/issues/37913 | [Bug]: Chrome extension relay: tab not found when using remote gateway + macOS node proxy (browser.proxy) | ### Bug type
Crash (process/app exits or hangs)
### Summary
The Chrome extension's tab registration message reaches the local relay at 127.0.0.1:18792, but when the relay is on a different machine than the gateway (local mode on Mac, remote gateway on VPS), the relay's tab list is not visible to the VPS browser.prox... | closed | completed | true | 2 | [
"bug",
"stale",
"bug:crash"
] | [] | 2026-03-06T14:32:30Z | 2026-04-27T04:55:00Z | 2026-04-24T22:05:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wikithoughts | 7,737,137 | MDQ6VXNlcjc3MzcxMzc= | User | false |
openclaw/openclaw | 4,034,598,742 | I_kwDOQb6kR87wexdW | 37,939 | https://github.com/openclaw/openclaw/issues/37939 | https://api.github.com/repos/openclaw/openclaw/issues/37939 | Telegram channel: stale-socket false positive every 30 minutes | ## Bug Description
The health monitor's stale-socket detection triggers false positives on Telegram channels every 30 minutes, causing unnecessary connection restarts and delivery failures.
## Root Cause
`DEFAULT_STALE_EVENT_THRESHOLD_MS = 30 * 60000` (30 minutes)
The health monitor was designed for Slack's WebSock... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T14:46:43Z | 2026-04-27T05:02:09Z | 2026-04-24T19:29:37Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | peter-claw | 260,595,439 | U_kgDOD4he7w | User | false |
openclaw/openclaw | 4,034,637,695 | I_kwDOQb6kR87we69_ | 37,943 | https://github.com/openclaw/openclaw/issues/37943 | https://api.github.com/repos/openclaw/openclaw/issues/37943 | [Bug]: Node Service fails to parse valid config on macOS | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
The `ai.openclaw.node` launchd service consistently fails to start because it cannot parse the main `~/.openclaw/openclaw.json` configuration file. It reports "Config invalid" in its logs, even when the configuration file has been verified ... | open | null | false | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-03-06T14:54:45Z | 2026-05-04T06:15:09Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | SicauxiaoqiangNo1 | 18,363,031 | MDQ6VXNlcjE4MzYzMDMx | User | false |
openclaw/openclaw | 4,034,670,080 | I_kwDOQb6kR87wfC4A | 37,955 | https://github.com/openclaw/openclaw/issues/37955 | https://api.github.com/repos/openclaw/openclaw/issues/37955 | session-memory hook: idle/daily session rollovers don't fire command:new/reset event | ## Summary
The bundled `session-memory` hook listens to `command:new` and `command:reset` events to save session context before a reset. This works correctly when a user explicitly types `/new` or `/reset`. However, sessions that expire via **idle timeout** or **daily reset** (the `session.reset.mode` policies) do NOT... | closed | not_planned | true | 3 | [
"stale"
] | [] | 2026-03-06T15:01:36Z | 2026-04-28T04:42:44Z | 2026-04-26T02:41:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | maertbot | 262,064,619 | U_kgDOD57J6w | User | false |
openclaw/openclaw | 4,034,698,883 | I_kwDOQb6kR87wfJ6D | 37,966 | https://github.com/openclaw/openclaw/issues/37966 | https://api.github.com/repos/openclaw/openclaw/issues/37966 | [Bug]: cacheRetention ignored for LiteLLM-proxied Anthropic models | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
`cacheRetention` configured in `agents.defaults.models` is silently ignored for LiteLLM-proxied Anthropic models (for example `litellm/claude-opus-4-6`).
No `cache_control` markers are injected into the API payload, so Anthropic prompt cac... | open | null | false | 5 | [] | [] | 2026-03-06T15:07:41Z | 2026-05-04T06:03:35Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | V0v1kkkAssistant | 260,059,260 | U_kgDOD4AwfA | User | false |
openclaw/openclaw | 4,034,758,759 | I_kwDOQb6kR87wfYhn | 37,983 | https://github.com/openclaw/openclaw/issues/37983 | https://api.github.com/repos/openclaw/openclaw/issues/37983 | Discord: forwarded text-only messages dropped by batching filter (baseText empty) | ## Problem
Forwarded text-only Discord messages (flag 16384, content in `message_snapshots`) are silently dropped when the user's own message text is empty. This happens even though `includeForwarded: true` is wired in the preflight handler.
## Root Cause
In the batching/flush logic (found in `reply-*.js`), there is... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-06T15:20:09Z | 2026-04-27T04:54:59Z | 2026-04-24T22:05:34Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | barneyontap | 257,155,530 | U_kgDOD1Phyg | User | false |
openclaw/openclaw | 4,034,816,897 | I_kwDOQb6kR87wfmuB | 38,038 | https://github.com/openclaw/openclaw/issues/38038 | https://api.github.com/repos/openclaw/openclaw/issues/38038 | [Feature Request] Agent Pack — shareable workspace templates for OpenClaw | ## Problem
Currently, OpenClaw has **Skills** (single capabilities) and **Plugins** (code extensions), both shareable via ClawHub/npm. But there's no way to share a **complete agent setup** — the combination of personality (SOUL.md), workspace config (AGENTS.md), multiple skills, and configuration that makes an agent ... | closed | not_planned | true | 2 | [
"stale"
] | [] | 2026-03-06T15:32:36Z | 2026-04-28T04:43:40Z | 2026-04-26T01:33:23Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | garysng | 84,123,412 | MDQ6VXNlcjg0MTIzNDEy | User | false |
openclaw/openclaw | 4,034,858,716 | I_kwDOQb6kR87wfw7c | 38,054 | https://github.com/openclaw/openclaw/issues/38054 | https://api.github.com/repos/openclaw/openclaw/issues/38054 | [Bug]: Windows install fails and immediately closes Powershell | ### Bug type
Crash (process/app exits or hangs)
### Summary
When trying to install OpenClaw on a FRESH Windows 11 install, the installer fails and immediately closes Powershell before allowing the user to read the reason for failure.
### Steps to reproduce
1. Start a Windows Sandbox
2. Open Powershell
3. Run `iwr ... | closed | completed | true | 7 | [
"bug",
"bug:crash"
] | [] | 2026-03-06T15:42:10Z | 2026-04-30T04:38:12Z | 2026-04-26T06:25:06Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | PwrSrg | 711,603 | MDQ6VXNlcjcxMTYwMw== | User | false |
openclaw/openclaw | 4,034,864,414 | I_kwDOQb6kR87wfyUe | 38,055 | https://github.com/openclaw/openclaw/issues/38055 | https://api.github.com/repos/openclaw/openclaw/issues/38055 | BUG: Subagent completion announce permanently lost when channel delivery is unavailable | ### Bug type
Regression (worked before, now fails)
### Summary
Subagent completion events are permanently dropped when the downstream channel (e.g., Telegram) is temporarily unreachable, because the announce pipeline is coupled to channel delivery. The event is never injected into the parent session, even though the... | closed | completed | true | 3 | [
"bug",
"stale",
"regression"
] | [] | 2026-03-06T15:43:31Z | 2026-04-27T05:02:08Z | 2026-04-24T19:29:48Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,034,874,045 | I_kwDOQb6kR87wf0q9 | 38,058 | https://github.com/openclaw/openclaw/issues/38058 | https://api.github.com/repos/openclaw/openclaw/issues/38058 | WhatsApp: messages lost during WebSocket reconnect — no retry after connection recovery | When the WhatsApp Web socket drops (503 / Stream Errored) and reconnects, any outbound messages that fail during the dead window are permanently lost. The gateway attempts retries against the closed connection but does not re-queue failed messages after the socket is re-established.
## Steps to reproduce
1. Agent gen... | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T15:45:38Z | 2026-04-27T05:02:07Z | 2026-04-24T19:29:52Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kimosahy | 260,060,902 | U_kgDOD4A25g | User | false |
openclaw/openclaw | 4,034,875,072 | I_kwDOQb6kR87wf07A | 38,059 | https://github.com/openclaw/openclaw/issues/38059 | https://api.github.com/repos/openclaw/openclaw/issues/38059 | Heartbeat 觸發後訊息跳過 (skipped) 未送到 Telegram | ## 版本
- OpenClaw: 2026.3.2
- 平台: Ubuntu 24.04
## 問題描述
- Heartbeat 已啟用 (30分鐘)
- 手動觸發 heartbeat 時，系統有產生 response preview (如 "晚安站長！🌙")
- 但最終狀態是 skipped，reason 是 "target-none"
- 訊息沒有送到 Telegram
## 期望行為
- Heartbeat 產生的訊息應該送到 Telegram
## 其他
- 正常對話沒有問題，只有 heartbeat 自動產生的訊息會被跳過 | closed | completed | true | 3 | [
"stale"
] | [] | 2026-03-06T15:45:51Z | 2026-04-27T05:02:07Z | 2026-04-24T19:29:56Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | mychataa | 128,360,829 | U_kgDOB6ahfQ | User | false |
openclaw/openclaw | 4,034,903,744 | I_kwDOQb6kR87wf77A | 38,066 | https://github.com/openclaw/openclaw/issues/38066 | https://api.github.com/repos/openclaw/openclaw/issues/38066 | Telegram: expose streamThrottleMs config for smoother native draft streaming | ## Problem
Native draft streaming via `sendMessageDraft` (added in v2026.3.2) works in DMs, but updates feel chunky — text appears in word bursts rather than smoothly character-by-character, like the native Telegram experience shown in [Durov's demo](https://x.com/durov/status/2028455440862830970).
The root cause is ... | closed | completed | false | 5 | [] | [] | 2026-03-06T15:52:10Z | 2026-05-03T14:29:04Z | 2026-05-03T14:29:04Z | NONE | null | 20260503T150031Z | 2026-05-03T15:00:31Z | SamuelZ12 | 68,740,129 | MDQ6VXNlcjY4NzQwMTI5 | User | false |
openclaw/openclaw | 4,034,227,990 | I_kwDOQb6kR87wdW8W | 37,844 | https://github.com/openclaw/openclaw/issues/37844 | https://api.github.com/repos/openclaw/openclaw/issues/37844 | [Bug]: Discord bot sends duplicate replies to the same message | ### Bug type
Regression (worked before, now fails)
### Summary
Discord bot sends duplicate replies to the same message (same message_id triggers two responses)
### Steps to reproduce
1. Configure Discord as a messaging channel
2. Send a message to the bot (direct message or group)
3. Observe that the bot sends two... | closed | completed | true | 7 | [
"bug",
"regression"
] | [] | 2026-03-06T13:26:19Z | 2026-04-27T05:02:15Z | 2026-04-24T19:28:57Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | lht0747 | 162,018,876 | U_kgDOCag2PA | User | false |
openclaw/openclaw | 4,292,748,556 | I_kwDOQb6kR87_3iUM | 69,158 | https://github.com/openclaw/openclaw/issues/69158 | https://api.github.com/repos/openclaw/openclaw/issues/69158 | [Bug]: spawn ENAMETOOLONG on Windows when using claude-cli provider | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
The claude-cli provider fails with `spawn ENAMETOOLONG` on every message on Windows. The Claude CLI works perfectly when called directly with prompts up to 32,000+ characters. The issue is that OpenClaw passes the prompt as ... | closed | completed | true | 2 | [
"bug",
"regression"
] | [] | 2026-04-20T02:31:43Z | 2026-04-28T05:16:55Z | 2026-04-25T12:29:26Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | cassioanorte | 106,702,931 | U_kgDOBlwoUw | User | false |
openclaw/openclaw | 4,293,143,607 | I_kwDOQb6kR87_5Cw3 | 69,196 | https://github.com/openclaw/openclaw/issues/69196 | https://api.github.com/repos/openclaw/openclaw/issues/69196 | Control UI microphone does not capture speech in Chrome on Mac mini | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The mic appears to activate for a split second, then stops with no transcript and no visible error.
### Steps to reproduce
Open OpenClaw Control UI in Chrome on macOS Mac mini at http://127.0.0.1:18789
Ensure... | closed | completed | true | 3 | [
"bug",
"bug:behavior"
] | [] | 2026-04-20T04:47:35Z | 2026-05-01T04:40:34Z | 2026-04-29T03:21:40Z | NONE | null | 20260501T113057Z | 2026-05-01T11:30:57Z | dmacelwain | 57,579,747 | MDQ6VXNlcjU3NTc5NzQ3 | User | false |
openclaw/openclaw | 4,022,477,859 | I_kwDOQb6kR87vwiQj | 34,574 | https://github.com/openclaw/openclaw/issues/34574 | https://api.github.com/repos/openclaw/openclaw/issues/34574 | loopDetection does not catch repeated exec tool calls | ## Bug: loopDetection does not catch repeated `exec` tool calls
### Environment
- OpenClaw version: v2026.3.2
- Model: Kimi K2.5 (via DashScope/Bailian, OpenAI-compatible API)
- Channel: webchat (control-ui)
### Description
The `tools.loopDetection` config fires for the `read` tool but **does not detect or block rep... | closed | completed | true | 9 | [] | [] | 2026-03-04T15:04:31Z | 2026-04-30T04:34:29Z | 2026-04-27T11:34:16Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gucasbrg | 3,182,658 | MDQ6VXNlcjMxODI2NTg= | User | false |
openclaw/openclaw | 4,293,218,488 | I_kwDOQb6kR87_5VC4 | 69,200 | https://github.com/openclaw/openclaw/issues/69200 | https://api.github.com/repos/openclaw/openclaw/issues/69200 | Feishu: message truncated when containing backticks or angle brackets in streaming mode (reopened) | ## Reopening #26466 and #26708
Both original issues were auto-closed and locked by the stale bot despite active fix PRs:
- **#26466** — Thinking content leaks to channel output when Think: off
- **#26708** — Message truncated when containing backticks or angle brackets in streaming mode
The bugs are still present and... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T05:07:25Z | 2026-04-30T04:38:55Z | 2026-04-26T05:25:19Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Vicky-v7 | 189,712,541 | U_kgDOC07InQ | User | false |
openclaw/openclaw | 4,293,250,730 | I_kwDOQb6kR87_5c6q | 69,202 | https://github.com/openclaw/openclaw/issues/69202 | https://api.github.com/repos/openclaw/openclaw/issues/69202 | Compaction failure leaves session in permanent failed state with no automatic recovery | ## Bug: Compaction failure leaves session in permanent failed state with no automatic recovery
### Summary
When session compaction fails (for any reason: timeout, API error, quota exceeded, model not supporting reasoning, etc.), the session enters `status=failed` permanently. There is no watchdog or automatic recover... | closed | completed | true | 2 | [] | [] | 2026-04-20T05:14:25Z | 2026-04-29T04:37:30Z | 2026-04-26T10:39:41Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | littlebest | 2,005,774 | MDQ6VXNlcjIwMDU3NzQ= | User | false |
openclaw/openclaw | 4,117,999,490 | I_kwDOQb6kR871c6-C | 52,613 | https://github.com/openclaw/openclaw/issues/52613 | https://api.github.com/repos/openclaw/openclaw/issues/52613 | [Bug]: Usage dashboard ignores archived session files (.jsonl.reset.*) | ### Bug type
Regression (worked before, now fails)
### Summary
Problem: The Usage page in the Control UI and /usage cost only aggregate data from active .jsonl session files. When sessions reset (via idle timeout, /new, etc.), the old transcript gets renamed to *.jsonl.reset.<timestamp> — and all usage data in those... | closed | completed | true | 3 | [
"bug",
"regression"
] | [] | 2026-03-23T03:42:16Z | 2026-04-30T04:48:05Z | 2026-04-25T06:02:33Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Og-Drill | 270,248,399 | U_kgDOEBupzw | User | false |
openclaw/openclaw | 4,293,274,966 | I_kwDOQb6kR87_5i1W | 69,205 | https://github.com/openclaw/openclaw/issues/69205 | https://api.github.com/repos/openclaw/openclaw/issues/69205 | Heartbeat configuration: no documented schema path or setup guide | ## Description
There is no documented way to configure heartbeat polling for agents. The feature is referenced in AGENTS.md templates and system prompts (e.g., `HEARTBEAT_OK` response pattern, heartbeat intervals), but there is no config schema path or setup documentation to actually enable it.
## Environment
- **Op... | closed | completed | true | 2 | [] | [] | 2026-04-20T05:19:59Z | 2026-04-28T04:49:36Z | 2026-04-25T19:05:15Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nekopoi | 17,149,318 | MDQ6VXNlcjE3MTQ5MzE4 | User | false |
openclaw/openclaw | 4,293,274,959 | I_kwDOQb6kR87_5i1P | 69,204 | https://github.com/openclaw/openclaw/issues/69204 | https://api.github.com/repos/openclaw/openclaw/issues/69204 | Cron isolated session agent-turn ends prematurely after 4.15 upgrade | ## Description
After upgrading from OpenClaw 4.11 to 4.15, cron jobs running in isolated sessions (`payload.kind: "agentTurn"`) consistently produce incomplete output. The agent loop appears to terminate prematurely — the model outputs transitional text like "Let me get more information with additional searches" and t... | closed | duplicate | true | 1 | [
"close:duplicate",
"dedupe:child"
] | [] | 2026-04-20T05:19:59Z | 2026-04-27T04:42:51Z | 2026-04-25T01:20:09Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | nekopoi | 17,149,318 | MDQ6VXNlcjE3MTQ5MzE4 | User | false |
openclaw/openclaw | 4,293,284,812 | I_kwDOQb6kR87_5lPM | 69,206 | https://github.com/openclaw/openclaw/issues/69206 | https://api.github.com/repos/openclaw/openclaw/issues/69206 | [Bug]: Telegram /dock_discord falls through to normal chat instead of native dock handling on 2026.4.15 | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
On OpenClaw 2026.4.15, Telegram /dock_discord is registered in the bot command menu but does not perform channel handoff and instead falls through to a normal chat/model response.
### Steps to reproduce
1. Ru... | closed | completed | true | 2 | [
"bug",
"bug:behavior"
] | [] | 2026-04-20T05:22:34Z | 2026-04-30T04:33:04Z | 2026-04-27T20:59:41Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | clawbones | 263,503,804 | U_kgDOD7S_vA | User | false |
openclaw/openclaw | 4,034,889,715 | I_kwDOQb6kR87wf4fz | 38,061 | https://github.com/openclaw/openclaw/issues/38061 | https://api.github.com/repos/openclaw/openclaw/issues/38061 | [Bug]: [Bug]: Webchat shows duplicate assistant messages because chat.history returns delivery-mirror transcript entries as normal assistant messages | ### Bug type
Regression (worked before, now fails)
### Summary
A single assistant reply is shown twice in OpenClaw webchat because chat.history returns both the real provider message and an internal delivery-mirror transcript entry, while Telegram only receives the message once.
### Steps to reproduce
1. Use OpenC... | closed | completed | true | 7 | [
"bug",
"regression"
] | [] | 2026-03-06T15:49:23Z | 2026-04-27T05:02:06Z | 2026-04-24T19:30:00Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | flyto168 | 204,641,331 | U_kgDODDKUMw | User | false |
openclaw/openclaw | 4,040,470,611 | I_kwDOQb6kR87w1LBT | 39,469 | https://github.com/openclaw/openclaw/issues/39469 | https://api.github.com/repos/openclaw/openclaw/issues/39469 | BUG: Control UI (webchat) double-records assistant messages in session JSONL | ### Bug type
Regression (worked before, now fails)
### Summary
The Control UI (webchat provider) records two assistant messages per LLM turn in the session JSONL file, causing duplicate responses in the conversation context and visible double-messages in the UI.
### Steps to reproduce
1. Start the OpenClaw gateway... | closed | completed | true | 6 | [
"bug",
"regression"
] | [] | 2026-03-08T04:42:56Z | 2026-04-27T04:59:46Z | 2026-04-24T20:31:30Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | kAIborg24 | 261,855,322 | U_kgDOD5uYWg | User | false |
openclaw/openclaw | 4,041,160,423 | I_kwDOQb6kR87w3zbn | 39,795 | https://github.com/openclaw/openclaw/issues/39795 | https://api.github.com/repos/openclaw/openclaw/issues/39795 | Telegram streaming "partial" creates duplicate messages due to generation counter race | ## Description
When using `streaming: "partial"` on Telegram channels, users see duplicate messages: the response appears twice briefly, then one copy is deleted. This happens consistently, especially on mobile where both notifications are visible.
## Environment
- OpenClaw version: `2026.3.2`
- Platform: Linux (WSL... | closed | completed | true | 5 | [] | [] | 2026-03-08T12:02:11Z | 2026-04-27T04:59:10Z | 2026-04-24T20:34:31Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | MarcoPambianchi | 130,920,816 | U_kgDOB82xcA | User | false |
openclaw/openclaw | 4,046,056,321 | I_kwDOQb6kR87xKeuB | 41,152 | https://github.com/openclaw/openclaw/issues/41152 | https://api.github.com/repos/openclaw/openclaw/issues/41152 | Async exec result delivery truncates output to ~400 chars and strips newlines | ## Bug Description
When exec commands go through the async approval flow (Discord exec approvals), the result delivered back to the agent via `System:` message injection is truncated to the **last ~400 characters** of output, and **all newlines are stripped**. The front of the output is silently dropped with no trunca... | open | null | false | 2 | [] | [] | 2026-03-09T14:49:27Z | 2026-05-04T09:13:53Z | null | NONE | null | 20260504T180039Z | 2026-05-04T18:00:39Z | Kaspre | 36,520,309 | MDQ6VXNlcjM2NTIwMzA5 | User | false |
openclaw/openclaw | 4,033,279,241 | I_kwDOQb6kR87wZvUJ | 37,584 | https://github.com/openclaw/openclaw/issues/37584 | https://api.github.com/repos/openclaw/openclaw/issues/37584 | feat: per-agent tool settings (e.g. tools.web.search.apiKey) | ## Feature Request
### Problem
In a multi-agent setup, `tools.web.search.apiKey` (and other tool-specific settings like provider, timeouts, etc.) are global — shared across all agents. There is no way to scope them per agent.
Per-agent `tools` config (`agents.list[].tools`) currently only supports:
- `profile` / `al... | open | null | false | 3 | [] | [] | 2026-03-06T09:51:11Z | 2026-05-04T06:03:19Z | null | NONE | null | 20260504T090105Z | 2026-05-04T09:01:05Z | ValentinH | 2,678,610 | MDQ6VXNlcjI2Nzg2MTA= | User | false |
openclaw/openclaw | 4,293,301,675 | I_kwDOQb6kR87_5pWr | 69,208 | https://github.com/openclaw/openclaw/issues/69208 | https://api.github.com/repos/openclaw/openclaw/issues/69208 | Umbrella: duplicate transcript, replay, and context assembly across channels | ## Summary
OpenClaw has a broader class of duplicate transcript, replay, and context assembly bugs that show up across multiple channels and runtime paths.
These bugs are currently reported as separate issues in MSTeams, webchat, Telegram, followup queue handling, delivery-mirror consumer paths, bootstrap reinjection... | open | null | false | 8 | [
"maintainer"
] | [] | 2026-04-20T05:26:25Z | 2026-05-02T03:04:03Z | null | MEMBER | null | 20260502T060142Z | 2026-05-02T06:01:42Z | BradGroux | 3,053,586 | MDQ6VXNlcjMwNTM1ODY= | User | false |
openclaw/openclaw | 4,219,846,176 | I_kwDOQb6kR877hb4g | 62,666 | https://github.com/openclaw/openclaw/issues/62666 | https://api.github.com/repos/openclaw/openclaw/issues/62666 | OpenClaw 4.5 — notifyActiveTaskWaiters TypeError crashes Discord gateway on incoming messages | ## Bug Description
After upgrading to OpenClaw 4.5 (2026.4.5), the gateway crashes with a TypeError when processing incoming Discord messages. The error originates in `notifyActiveTaskWaiters` in `command-queue.ts` — it attempts to iterate over an undefined value (expecting an array). This crashes the gateway handler ... | closed | completed | true | 0 | [] | [] | 2026-04-07T18:40:13Z | 2026-04-23T04:22:40Z | 2026-04-20T05:46:44Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | mentholmike | 68,653,554 | MDQ6VXNlcjY4NjUzNTU0 | User | false |
openclaw/openclaw | 4,293,392,074 | I_kwDOQb6kR87_5_bK | 69,214 | https://github.com/openclaw/openclaw/issues/69214 | https://api.github.com/repos/openclaw/openclaw/issues/69214 | [Bug]: Gateway client gets stuck in scope-upgrade repair loop for Telegram Native Approvals | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
Telegram Native Approvals gets stuck in a scope-upgrade pairing repair loop and cannot complete re-pairing. Each approval attempt causes a new pending repair request to be generated, so the upgrade never settle... | closed | completed | true | 2 | [
"bug"
] | [] | 2026-04-20T05:48:05Z | 2026-04-28T04:49:35Z | 2026-04-25T19:05:22Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | legiang2907214995 | 274,598,721 | U_kgDOEF4LQQ | User | false |
openclaw/openclaw | 4,293,408,153 | I_kwDOQb6kR87_6DWZ | 69,216 | https://github.com/openclaw/openclaw/issues/69216 | https://api.github.com/repos/openclaw/openclaw/issues/69216 | openclaw doctor and upgrade should backup .env and openclaw.json before modifying | ## Description
`openclaw doctor` (and `openclaw update`) can overwrite `.env` and `openclaw.json` without creating a backup first. This can result in loss of user-configured environment variables.
## Environment
- **OpenClaw version at time of incident**: Upgrading from 2026.4.10 to 2026.4.11
- **OS**: Ubuntu 24.04,... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T05:51:49Z | 2026-04-28T04:49:35Z | 2026-04-25T19:05:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | nekopoi | 17,149,318 | MDQ6VXNlcjE3MTQ5MzE4 | User | false |
openclaw/openclaw | 4,293,435,510 | I_kwDOQb6kR87_6KB2 | 69,220 | https://github.com/openclaw/openclaw/issues/69220 | https://api.github.com/repos/openclaw/openclaw/issues/69220 | Gemini text-tag reasoning conflicts with native thinking — produces unclosed <think>, empty post-tool turn, payloads=0 | ## Summary
OpenClaw 2026.4.14 (and prior 2026.4.x) hardcodes `google-generative-ai` to **tagged** reasoning output mode in `dist/provider-utils-CWYcvxhN.js`:
```js
const BUILTIN_REASONING_OUTPUT_MODES = { "google-generative-ai": "tagged" };
```
`tagged` mode injects this directive into the system prompt (via `dist/s... | open | null | false | 3 | [] | [] | 2026-04-20T05:58:05Z | 2026-05-02T16:55:34Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | mrbrl | 2,474,918 | MDQ6VXNlcjI0NzQ5MTg= | User | false |
openclaw/openclaw | 4,293,534,495 | I_kwDOQb6kR87_6iMf | 69,224 | https://github.com/openclaw/openclaw/issues/69224 | https://api.github.com/repos/openclaw/openclaw/issues/69224 | v2026.4.15 pnpm install ENOENT: patches/ dir missing from published image | ## Summary
Building a downstream image `FROM ghcr.io/openclaw/openclaw:2026.4.15` and running `pnpm install` inside `/app` fails with `ENOENT` because `package.json` references a patched dependency whose patch file is not present in the published image.
## Reproduction
```
$ docker run --rm --entrypoint sh ghcr.io/o... | closed | completed | true | 1 | [] | [
"vincentkoc"
] | 2026-04-20T06:22:26Z | 2026-04-30T04:38:46Z | 2026-04-26T05:27:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | gucasbrg | 3,182,658 | MDQ6VXNlcjMxODI2NTg= | User | false |
openclaw/openclaw | 4,293,551,056 | I_kwDOQb6kR87_6mPQ | 69,229 | https://github.com/openclaw/openclaw/issues/69229 | https://api.github.com/repos/openclaw/openclaw/issues/69229 | [Bug]: tasks audit reports false-positive inconsistent_timestamps on succeeded/failed/running tasks | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
## Summary
`openclaw tasks audit` is reporting a large number of `inconsistent_timestamps` warnings with:
> startedAt is earlier than createdAt
This appears on tasks with multiple outcomes, including:
- `succeeded`
- `fai... | closed | completed | true | 5 | [
"bug",
"regression"
] | [] | 2026-04-20T06:26:26Z | 2026-04-29T04:42:45Z | 2026-04-26T06:57:52Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | aemeathpa | 269,386,194 | U_kgDOEA6B0g | User | false |
openclaw/openclaw | 4,293,580,656 | I_kwDOQb6kR87_6tdw | 69,234 | https://github.com/openclaw/openclaw/issues/69234 | https://api.github.com/repos/openclaw/openclaw/issues/69234 | [Feature]: Allow compaction summarization base instructions to be configured via openclaw.json | ## Problem
The base summarization instructions used during context compaction are hardcoded in `src/agents/compaction.ts`:
1. `MERGE_SUMMARIES_INSTRUCTIONS` — the "MUST PRESERVE" template telling the model what to keep in summaries
2. `IDENTIFIER_PRESERVATION_INSTRUCTIONS` — the "Preserve all opaque identifiers..." t... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T06:33:14Z | 2026-04-28T04:49:33Z | 2026-04-25T19:05:41Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | AxelHu | 12,809,280 | MDQ6VXNlcjEyODA5Mjgw | User | false |
openclaw/openclaw | 4,293,591,762 | I_kwDOQb6kR87_6wLS | 69,235 | https://github.com/openclaw/openclaw/issues/69235 | https://api.github.com/repos/openclaw/openclaw/issues/69235 | [Bug]: Opus / Sonnet reasoning leakage (on Slack) | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
This behavior has been around for a few versions. It looks like the tagging for reasoning doesn't always work properly on Sonnet 4.6 / Opus 4.6 (didn't test on 4.7).
This leads to reasoning (internal monologue... | closed | completed | true | 1 | [
"bug"
] | [] | 2026-04-20T06:35:31Z | 2026-04-30T04:38:18Z | 2026-04-26T06:16:08Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | oferlaor | 3,070,108 | MDQ6VXNlcjMwNzAxMDg= | User | false |
openclaw/openclaw | 4,293,573,037 | I_kwDOQb6kR87_6rmt | 69,233 | https://github.com/openclaw/openclaw/issues/69233 | https://api.github.com/repos/openclaw/openclaw/issues/69233 | [Feature]: WhatsApp Group Admin Privacy Feature | ### Summary
Add admin designation for WhatsApp groups to prevent unauthorized discovery of bot capabilities by restricting non-admin group members' access to commands and mentions.
### Problem to solve
Currently, all allowlisted users in a WhatsApp group can interact with the bot, potentially discovering its capabil... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-20T06:31:34Z | 2026-04-30T04:38:54Z | 2026-04-26T05:25:23Z | CONTRIBUTOR | null | 20260430T060018Z | 2026-04-30T06:00:18Z | haishmg | 4,529,977 | MDQ6VXNlcjQ1Mjk5Nzc= | User | false |
openclaw/openclaw | 4,078,749,613 | I_kwDOQb6kR87zHMet | 47,499 | https://github.com/openclaw/openclaw/issues/47499 | https://api.github.com/repos/openclaw/openclaw/issues/47499 | [Feature]: Dashboard Embedded VNC Viewer + Agent Full Desktop Takeover (Manus-like, evolving to full computer control) | ### Summary
<!-- Failed to upload "lll_high.gif" -->
### [Feature Request] Dashboard Embedded VNC Viewer + Agent Full Desktop Takeover (Manus-like, evolving to full computer control)
#### Summary
I have already implemented a **fully functional VNC viewer** inside the OpenClaw Dashboard. It runs very well in the brow... | open | null | false | 2 | [
"enhancement"
] | [] | 2026-03-15T16:55:43Z | 2026-05-04T01:32:54Z | null | NONE | null | 20260504T030031Z | 2026-05-04T03:00:31Z | dapangyu-fish | 68,261,559 | MDQ6VXNlcjY4MjYxNTU5 | User | false |
openclaw/openclaw | 4,293,658,054 | I_kwDOQb6kR87_7AXG | 69,238 | https://github.com/openclaw/openclaw/issues/69238 | https://api.github.com/repos/openclaw/openclaw/issues/69238 | Architecture: remove duplicate BlueBubbles schema in core (boundary violation) | ## Summary
Core (`src/config/zod-schema.providers-core.ts:1425-1492`) defines its own `BlueBubblesAccountSchemaBase` / `BlueBubblesAccountSchema` / `BlueBubblesConfigSchema` that duplicates the plugin-owned schema in `extensions/bluebubbles/src/config-schema.ts`. Schema additions need to be made in **two** places to l... | open | null | false | 0 | [
"maintainer"
] | [
"omarshahine"
] | 2026-04-20T06:48:44Z | 2026-04-20T06:48:53Z | null | CONTRIBUTOR | null | 20260420T075759Z | 2026-04-20T07:57:59Z | omarshahine | 10,343,873 | MDQ6VXNlcjEwMzQzODcz | User | false |
openclaw/openclaw | 4,293,734,338 | I_kwDOQb6kR87_7S_C | 69,241 | https://github.com/openclaw/openclaw/issues/69241 | https://api.github.com/repos/openclaw/openclaw/issues/69241 | GitHub Copilot provider: add Claude Opus 4.7 to default model list | ## Problem
The GitHub Copilot provider's `DEFAULT_MODEL_IDS` in `extensions/github-copilot/models-defaults.ts` does not include Claude Opus models. The current hardcoded list is:
```js
const DEFAULT_MODEL_IDS = [
"claude-sonnet-4.6",
"claude-sonnet-4.5",
"gpt-4o",
"gpt-4.1",
"gpt-4.1-mini",
"gpt-4.1-nano"... | closed | completed | true | 1 | [] | [] | 2026-04-20T07:04:10Z | 2026-04-28T04:49:31Z | 2026-04-25T19:05:54Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | kiote | 123,637 | MDQ6VXNlcjEyMzYzNw== | User | false |
openclaw/openclaw | 4,293,735,131 | I_kwDOQb6kR87_7TLb | 69,242 | https://github.com/openclaw/openclaw/issues/69242 | https://api.github.com/repos/openclaw/openclaw/issues/69242 | exec tool on Linux intermittently SIGKILLs broad find/grep discovery commands without OOM evidence | ## Summary
On Linux, broad file-system search commands run via the OpenClaw `exec` tool are repeatedly terminated with `SIGKILL` / `terminated`, even though normal narrow commands and file `read`/`edit` flows continue to work.
This does **not** currently look like a kernel OOM kill on the host. `journalctl -k` did not... | open | null | false | 2 | [] | [] | 2026-04-20T07:04:21Z | 2026-05-02T12:54:01Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | schksw-ctrl | 258,609,409 | U_kgDOD2oRAQ | User | false |
openclaw/openclaw | 4,293,684,335 | I_kwDOQb6kR87_7Gxv | 69,239 | https://github.com/openclaw/openclaw/issues/69239 | https://api.github.com/repos/openclaw/openclaw/issues/69239 | [Bug] Image tool silently drops all custom models when any provider lacks apiKey | ## Description
The `image` tool fails to find ANY custom models (e.g. `trapi/K2.6-code-preview`) with error "Unknown model", even though the model exists in `models.json` with correct `input: ["text", "image"]`.
## Root Cause
The image tool uses `PiModelRegistry` from `@mariozechner/pi-coding-agent` to load models v... | closed | completed | true | 3 | [] | [] | 2026-04-20T06:53:56Z | 2026-04-28T04:49:32Z | 2026-04-25T19:05:48Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | TNTest | 470,968 | MDQ6VXNlcjQ3MDk2OA== | User | false |
openclaw/openclaw | 4,293,537,396 | I_kwDOQb6kR87_6i50 | 69,225 | https://github.com/openclaw/openclaw/issues/69225 | https://api.github.com/repos/openclaw/openclaw/issues/69225 | Add email ingestion to heartbeat digest — surface useful emails, filter noise | ## Context
OpenClaw's heartbeat digest (as used in the Cleo setup at altx-labs/openclaw) currently ingests:
- WhatsApp + Telegram group messages
- News (via Exa)
- Blogs (via blogwatcher RSS)
- X/Twitter (via Grok + `x_search`)
Email is the obvious missing channel. A huge fraction of actionable signal lives in inbox ... | closed | not_planned | true | 2 | [] | [] | 2026-04-20T06:23:05Z | 2026-04-28T04:49:34Z | 2026-04-25T19:05:35Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | wiiiimm | 179,761 | MDQ6VXNlcjE3OTc2MQ== | User | false |
openclaw/openclaw | 4,293,770,463 | I_kwDOQb6kR87_7bzf | 69,246 | https://github.com/openclaw/openclaw/issues/69246 | https://api.github.com/repos/openclaw/openclaw/issues/69246 | [Bug]: OpenClaw website keeps loading indefinitely instead of showing custom MCP install errors | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When installing a custom MCP, the OpenClaw website keeps showing an infinite loading state instead of surfacing the actual error. The real error is only visible in the OpenClaw TUI. This makes it hard to unders... | closed | completed | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-20T07:11:30Z | 2026-04-28T04:49:31Z | 2026-04-25T19:06:07Z | CONTRIBUTOR | null | 20260428T060020Z | 2026-04-28T06:00:20Z | skainguyen1412 | 14,249,881 | MDQ6VXNlcjE0MjQ5ODgx | User | false |
openclaw/openclaw | 4,163,860,426 | I_kwDOQb6kR874L3fK | 56,962 | https://github.com/openclaw/openclaw/issues/56962 | https://api.github.com/repos/openclaw/openclaw/issues/56962 | /compact command not working in Control UI | **环境：**
- OpenClaw version: 2026.3.24 (cff6dc9)
- Client: Control UI (webchat)
- OS: macOS (Darwin 25.3.0 arm64)
**问题：**
在 Control UI 聊天输入框输入 `/compact` 后没有任何反应。命令不会被发送到 Gateway。
**验证：**
1. Gateway 日志中没有任何 `/compact` 相关的请求记录
2. 其他命令如 `/reset`、`/new` 工作正常
3. 设备已正确配对，包含 `operator.admin` scope
4. 重启 Gateway 和刷新页面都无法解决
... | closed | completed | true | 4 | [] | [] | 2026-03-29T11:20:16Z | 2026-04-28T05:19:49Z | 2026-04-25T11:53:58Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | TTwms | 40,107,518 | MDQ6VXNlcjQwMTA3NTE4 | User | false |
openclaw/openclaw | 4,293,846,550 | I_kwDOQb6kR87_7uYW | 69,249 | https://github.com/openclaw/openclaw/issues/69249 | https://api.github.com/repos/openclaw/openclaw/issues/69249 | Gateway restart can silently abort an in-flight Discord turn, with no automatic recovery message to the user | ### Summary
When the OpenClaw gateway restarts during an active Discord-backed turn, the in-flight reply can abort, but the user may receive no visible recovery message in the channel. From the user perspective, the assistant appears to ghost until they manually ping again.
### Environment
- OpenClaw version: `OpenC... | open | null | false | 1 | [] | [] | 2026-04-20T07:23:22Z | 2026-05-02T12:13:03Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | slideshow-dingo | 258,996,445 | U_kgDOD2_43Q | User | false |
openclaw/openclaw | 4,293,864,650 | I_kwDOQb6kR87_7yzK | 69,250 | https://github.com/openclaw/openclaw/issues/69250 | https://api.github.com/repos/openclaw/openclaw/issues/69250 | Gateway calls plugin register() ~30x per boot for non-bundled plugins (CostClaw thrash) | ## Summary
On a single gateway process boot, the plugin loader invokes a non-bundled plugin's `register(api)` default export approximately **30 times in rapid succession** (sub-second spacing), despite the loader having a `seenIds` Map dedup. Combined with slow plugin init paths (e.g. `better-sqlite3` native binding +... | closed | completed | true | 1 | [] | [] | 2026-04-20T07:26:09Z | 2026-04-27T04:48:33Z | 2026-04-24T23:21:16Z | CONTRIBUTOR | null | 20260427T090020Z | 2026-04-27T09:00:20Z | markthebest12 | 6,416,874 | MDQ6VXNlcjY0MTY4NzQ= | User | false |
openclaw/openclaw | 4,293,866,252 | I_kwDOQb6kR87_7zMM | 69,251 | https://github.com/openclaw/openclaw/issues/69251 | https://api.github.com/repos/openclaw/openclaw/issues/69251 | exec host=node SYSTEM_RUN_DENIED for all file-referencing commands due to /bin/sh -lc wrapping | ## Version
`2026.4.14`
## Platform
macOS (arm64), paired node connecting via Tailscale
## Summary
Any `exec host=node` command that references a file existing on the remote node is denied with `SYSTEM_RUN_DENIED: approval cannot safely bind this interpreter/runtime command`, even with `security: "full"` and `ask: "o... | closed | not_planned | true | 2 | [] | [] | 2026-04-20T07:26:23Z | 2026-04-29T04:33:27Z | 2026-04-27T00:01:43Z | CONTRIBUTOR | null | 20260429T060025Z | 2026-04-29T06:00:25Z | Blakeshannon | 257,822,860 | U_kgDOD14QjA | User | false |
openclaw/openclaw | 4,293,902,564 | I_kwDOQb6kR87_78Dk | 69,254 | https://github.com/openclaw/openclaw/issues/69254 | https://api.github.com/repos/openclaw/openclaw/issues/69254 | Bug: IM channels like Telegram/QQ do not get recovery-aware gateway restart behavior | ## Summary
Gateway restart recovery behavior appears to diverge by chat run surface. In current testing, **Control UI/WebChat gets recovery-aware restart behavior, while IM channels such as Telegram direct do not**.
The practical result is that after a gateway restart, Control UI sessions can recover through the inte... | closed | completed | true | 1 | [] | [] | 2026-04-20T07:32:02Z | 2026-04-28T04:49:30Z | 2026-04-25T19:06:21Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | Silronin | 67,084,317 | MDQ6VXNlcjY3MDg0MzE3 | User | false |
openclaw/openclaw | 4,293,906,231 | I_kwDOQb6kR87_7883 | 69,255 | https://github.com/openclaw/openclaw/issues/69255 | https://api.github.com/repos/openclaw/openclaw/issues/69255 | [Bug]: Feishu WebSocket fails to reconnect after token timeout during gateway restart/bootstrap | ### Bug type
Behavior bug (incorrect output/state without crash)
### Summary
飞书 `tenant_access_token` 获取请求超时（尤其在网关重启/启动阶段）后，飞书 WebSocket 连接无法自动恢复，对话中断。
### Steps to reproduce
1. 配置飞书插件，使用 `connectionMode: "websocket"`。
2. 重启网关（或关机重启后重新启动）。
3. 启动时若 `open.feishu.cn` 响应慢（超时 30s），日志会出现：
```
[model-pricing] prici... | closed | not_planned | true | 3 | [] | [] | 2026-04-20T07:32:39Z | 2026-04-28T04:49:29Z | 2026-04-25T19:06:28Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | xiep-dot | 266,485,161 | U_kgDOD-I9qQ | User | false |
openclaw/openclaw | 4,294,015,988 | I_kwDOQb6kR87_8Xv0 | 69,260 | https://github.com/openclaw/openclaw/issues/69260 | https://api.github.com/repos/openclaw/openclaw/issues/69260 | Harden Gemini ACP integration against ambient API-key fallback and add generic auth-contract hooks for ACP agents | ### Summary
This is an integration-hardening / defense-in-depth issue, not a claim that OpenClaw is the sole root cause of Gemini auth drift.
We just finished a local incident/debugging cycle where a risky Gemini ACP path could drift durable Gemini auth away from `oauth-personal` and into the API-key lane when `GEMINI... | open | null | false | 1 | [] | [] | 2026-04-20T07:52:04Z | 2026-05-02T16:41:03Z | null | CONTRIBUTOR | null | 20260502T180029Z | 2026-05-02T18:00:29Z | anyech | 8,743,351 | MDQ6VXNlcjg3NDMzNTE= | User | false |
openclaw/openclaw | 4,158,361,582 | I_kwDOQb6kR87324_u | 56,262 | https://github.com/openclaw/openclaw/issues/56262 | https://api.github.com/repos/openclaw/openclaw/issues/56262 | 2026.3.24 breaks Mistral provider: models.json overwritten with wrong api type | ## Description
Upgrading from 2026.3.13 to 2026.3.24 breaks the Mistral provider. On gateway startup, 2026.3.24 overwrites `agents/main/agent/models.json` and adds a Mistral provider entry with `"api": "openai-completions"`. This causes all Mistral API calls to return a 404:
```json
{
"message": "no Route matched w... | closed | completed | true | 3 | [] | [] | 2026-03-28T07:37:42Z | 2026-04-29T04:40:53Z | 2026-04-26T08:40:53Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | samgoos7 | 139,918,151 | U_kgDOCFb7Rw | User | false |
openclaw/openclaw | 4,294,072,499 | I_kwDOQb6kR87_8liz | 69,264 | https://github.com/openclaw/openclaw/issues/69264 | https://api.github.com/repos/openclaw/openclaw/issues/69264 | Clarification: Does OpenClaw Support Dynamic Agent Updates Without Gateway Restart? | ### Summary
Clarify which OpenClaw configurations (agents, routing, channels) support hot reload and which require a gateway restart.
### Problem to solve
When building a multi-agent, multi-user system, it's unclear which configuration changes can be applied dynamically and which require restarting the OpenClaw Gate... | closed | completed | true | 0 | [
"enhancement"
] | [] | 2026-04-20T08:02:26Z | 2026-04-23T04:22:38Z | 2026-04-20T08:02:36Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | F32138 | 145,747,124 | U_kgDOCK_stA | User | false |
openclaw/openclaw | 4,294,083,819 | I_kwDOQb6kR87_8oTr | 69,265 | https://github.com/openclaw/openclaw/issues/69265 | https://api.github.com/repos/openclaw/openclaw/issues/69265 | [Feature]: Clarification: Does OpenClaw Support Dynamic Agent Updates Without Gateway Restart? | ### Summary
Clarify which OpenClaw configurations (agents, routing, channels) support hot reload and which require a gateway restart.
### Problem to solve
When building a multi-agent, multi-user system, it's unclear which configuration changes can be applied dynamically and which require restarting the OpenClaw Gat... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-04-20T08:04:32Z | 2026-04-28T04:49:28Z | 2026-04-25T19:06:48Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | F32138 | 145,747,124 | U_kgDOCK_stA | User | false |
openclaw/openclaw | 4,294,270,019 | I_kwDOQb6kR87_9VxD | 69,271 | https://github.com/openclaw/openclaw/issues/69271 | https://api.github.com/repos/openclaw/openclaw/issues/69271 | Add per-session delete/archive action button in Control UI Sessions view | Currently, deleting sessions in the Control UI requires selecting checkboxes and using the bulk delete bar. This works for batch operations but is cumbersome for quick single-session actions.
**Proposed changes:**
1. **Per-row action button** — Add a 🗑️ (delete) and/or 📦 (archive) icon button at the end of each ses... | open | null | false | 1 | [] | [] | 2026-04-20T08:34:23Z | 2026-05-02T13:08:02Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | geek111 | 6,862,522 | MDQ6VXNlcjY4NjI1MjI= | User | false |
openclaw/openclaw | 4,294,339,814 | I_kwDOQb6kR87_9mzm | 69,274 | https://github.com/openclaw/openclaw/issues/69274 | https://api.github.com/repos/openclaw/openclaw/issues/69274 | Qwen: qwen3.6-plus is blocked on Coding Plan Pro endpoint despite being officially supported | ## Problem
OpenClaw hardcodes a filter that blocks `qwen3.6-plus` when the configured Qwen provider uses the Coding Plan endpoint (`https://coding.dashscope.aliyuncs.com/v1`), even though Alibaba Cloud officially includes `qwen3.6-plus` as the **recommended flagship model** for the Coding Plan Pro subscription.
---
... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T08:45:06Z | 2026-04-30T04:38:53Z | 2026-04-26T05:25:27Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | Elvisvon | 135,479,820 | U_kgDOCBNCDA | User | false |
openclaw/openclaw | 4,294,282,502 | I_kwDOQb6kR87_9Y0G | 69,272 | https://github.com/openclaw/openclaw/issues/69272 | https://api.github.com/repos/openclaw/openclaw/issues/69272 | feat(agents): forward prompt_cache_key on openai-completions transport when compat.supportsPromptCacheKey is set | ## Summary
`buildOpenAIResponsesParams` forwards `prompt_cache_key: options?.sessionId` on every Responses API turn (`src/agents/openai-transport-stream.ts:794`), and `parseTransportChunkUsage` already reads `prompt_tokens_details.cached_tokens` on Completions responses (`src/agents/openai-transport-stream.ts:1546`).
... | closed | completed | true | 2 | [] | [] | 2026-04-20T08:36:22Z | 2026-04-27T04:38:46Z | 2026-04-25T02:52:03Z | NONE | null | 20260427T090020Z | 2026-04-27T09:00:20Z | wallscaler | 7,602,667 | MDQ6VXNlcjc2MDI2Njc= | User | false |
openclaw/openclaw | 4,294,502,626 | I_kwDOQb6kR87_-Oji | 69,279 | https://github.com/openclaw/openclaw/issues/69279 | https://api.github.com/repos/openclaw/openclaw/issues/69279 | Feature request: Kokoro TTS as native provider | ## Feature Request: Kokoro TTS as native provider
**Fecha:** 2026-04-20
**Versión OpenClaw:** 2026.4.15
## Descripción
Solicito considerar añadir **Kokoro TTS** (hexgrad/Kokoro-82M) como proveedor nativo de TTS en OpenClaw.
## Contexto
Kokoro es un modelo TTS open-source (Apache 2.0) con 82M parámetros:
- Alta cal... | closed | not_planned | true | 1 | [] | [] | 2026-04-20T09:11:27Z | 2026-04-28T04:49:28Z | 2026-04-25T19:06:55Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | urrielu | 213,940,341 | U_kgDODMB4dQ | User | false |
openclaw/openclaw | 3,883,427,641 | I_kwDOQb6kR87neGc5 | 6,731 | https://github.com/openclaw/openclaw/issues/6731 | https://api.github.com/repos/openclaw/openclaw/issues/6731 | [Feature]: safe/unsafe ClawdBot | perhaps inherit this feature or abstraction from Rust, completely rewrite this project in Rust. Safe mode can be running ClawdBot locally in a sandbox environment with limited access and it protects user from undefined behaviour, memory leaks and root access from who knows outside, and unsafe it will be given root perm... | open | null | false | 12 | [
"enhancement"
] | [] | 2026-02-02T00:43:44Z | 2026-05-05T00:07:55Z | null | NONE | null | 20260505T030055Z | 2026-05-05T03:00:55Z | ChrisX101010 | 143,946,165 | U_kgDOCJRxtQ | User | false |
openclaw/openclaw | 4,294,544,494 | I_kwDOQb6kR87_-Yxu | 69,280 | https://github.com/openclaw/openclaw/issues/69280 | https://api.github.com/repos/openclaw/openclaw/issues/69280 | compaction.truncateAfterCompaction rejected by strict zod schema | ## Summary
`truncateAfterCompaction` is declared in the TypeScript type (`types.agent-defaults.ts:451`), read at runtime (`compact.ts:1103`), documented in schema help (`schema.help.ts:1216`), and has a label (`schema.labels.ts:564`) — but is **missing from the zod validation schema** at `zod-schema.agent-defaults.ts:... | closed | completed | true | 1 | [] | [] | 2026-04-20T09:18:44Z | 2026-04-25T04:17:35Z | 2026-04-22T16:59:29Z | CONTRIBUTOR | null | 20260425T060021Z | 2026-04-25T06:00:21Z | sukhdeepjohar | 46,626,869 | MDQ6VXNlcjQ2NjI2ODY5 | User | false |
openclaw/openclaw | 4,294,553,062 | I_kwDOQb6kR87_-a3m | 69,281 | https://github.com/openclaw/openclaw/issues/69281 | https://api.github.com/repos/openclaw/openclaw/issues/69281 | Bug: Cron jobs with delivery.mode: "none" incorrectly show "not-delivered" error | # Bug: Cron jobs with `delivery.mode: "none"` incorrectly show "not-delivered" error
## Summary
When a cron job is configured with `delivery: { mode: "none" }` to explicitly opt out of message delivery, the system still records:
- `deliveryStatus: "not-delivered"` (should be "not-requested")
- `deliveryError: "⚠️ ✉️ ... | closed | completed | true | 0 | [] | [] | 2026-04-20T09:20:13Z | 2026-04-23T04:22:36Z | 2026-04-20T13:15:55Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | matsuri1987 | 118,504,844 | U_kgDOBxA9jA | User | false |
openclaw/openclaw | 4,294,701,834 | I_kwDOQb6kR87_-_MK | 69,284 | https://github.com/openclaw/openclaw/issues/69284 | https://api.github.com/repos/openclaw/openclaw/issues/69284 | Gateway recurring pairing required error - subagent sessions fail | ## Bug: Gateway recurring "pairing required" error
### Description
OpenClaw gateway repeatedly shows "gateway closed (1008): pairing required" error, preventing subagent sessions from starting. This happens even for local connections from 127.0.0.1.
### Environment
- OpenClaw version: 2026.4.14
- OS: Linux
- Gateway ... | closed | duplicate | true | 1 | [
"close:duplicate"
] | [] | 2026-04-20T09:43:55Z | 2026-04-23T04:21:58Z | 2026-04-21T03:26:23Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | Pastafarian23 | 10,189,603 | MDQ6VXNlcjEwMTg5NjAz | User | false |
openclaw/openclaw | 4,294,768,637 | I_kwDOQb6kR87__Pf9 | 69,287 | https://github.com/openclaw/openclaw/issues/69287 | https://api.github.com/repos/openclaw/openclaw/issues/69287 | Bug: Session token count not reset after compaction | # Bug: Session token count not reset after compaction
## Summary
After compaction clears messages, `session.totalTokens` is not reset to 0, causing the compaction safeguard to trigger repeatedly in an infinite loop.
## Environment
- **OpenClaw Version**: 2026.4.15 (041266a)
- **OS**: macOS (Darwin 25.3.0 arm64)
- **N... | closed | not_planned | true | 3 | [] | [] | 2026-04-20T09:55:35Z | 2026-04-30T04:38:52Z | 2026-04-26T05:25:36Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | RrcioElise | 47,847,992 | MDQ6VXNlcjQ3ODQ3OTky | User | false |
openclaw/openclaw | 4,294,789,378 | I_kwDOQb6kR87__UkC | 69,289 | https://github.com/openclaw/openclaw/issues/69289 | https://api.github.com/repos/openclaw/openclaw/issues/69289 | [Bug]: Browser aria snapshot refs resolve in snapshots but fail in follow-up actions | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
In my local OpenClaw runtime, browser page open/snapshot works, but follow-up browser actions against aria snapshot refs fail with locator timeouts.
### Steps to reproduce
Minimal repro
1. Open https://example.com
2. Take... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-20T09:58:57Z | 2026-04-24T04:25:54Z | 2026-04-22T02:09:28Z | NONE | null | 20260424T060021Z | 2026-04-24T06:00:21Z | FoundryHouseAI | 271,721,896 | U_kgDOEDIlqA | User | false |
openclaw/openclaw | 4,294,861,252 | I_kwDOQb6kR87__mHE | 69,291 | https://github.com/openclaw/openclaw/issues/69291 | https://api.github.com/repos/openclaw/openclaw/issues/69291 | [Feature]: [Proposal] Add systematic Agent Behavior Principles to system prompt | ### Summary
Add systematic Agent Behavior Principles (Think Before Acting, Shortest Path, Surgical Operations, Verify Outcomes) as a dedicated component in the system prompt
### Problem to solve
OpenClaw's `rules.ts` contains scattered behavioral constraints (e.g., "do not assume success", "use tools efficiently"), ... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-20T10:10:11Z | 2026-04-29T04:33:27Z | 2026-04-27T00:01:48Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | dongangL-art | 267,890,659 | U_kgDOD_ev4w | User | false |
openclaw/openclaw | 4,294,888,102 | I_kwDOQb6kR87__sqm | 69,293 | https://github.com/openclaw/openclaw/issues/69293 | https://api.github.com/repos/openclaw/openclaw/issues/69293 | [Bug]: zsh completion for openclaw wiki ingest does not complete file paths | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
The generated zsh completion for `openclaw wiki ingest` knows about the command, but omits the positional file argument, so path completion does not work.
### Steps to reproduce
## Environment
- OpenClaw ver... | closed | not_planned | true | 1 | [
"bug",
"bug:behavior"
] | [] | 2026-04-20T10:14:17Z | 2026-04-29T04:36:28Z | 2026-04-26T12:05:30Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | hombrevago | 134,784,400 | U_kgDOCAilkA | User | false |
openclaw/openclaw | 4,294,895,247 | I_kwDOQb6kR87__uaP | 69,294 | https://github.com/openclaw/openclaw/issues/69294 | https://api.github.com/repos/openclaw/openclaw/issues/69294 | [Bug]: ACP plugin crashes gateway on Windows 11 - embedded ACP runtime probe failed | ### Bug type
Crash (process/app exits or hangs)
### Beta release blocker
No
### Summary
On a fresh install of OpenClaw on Windows 11 (HP Elitebook 840 G9), the gateway crashes immediately due to "embedded ACP runtime probe failed (exit=3221225783,signal=null)". Disabling the plugin via config does not prevent it ... | closed | completed | true | 1 | [
"bug",
"bug:crash"
] | [] | 2026-04-20T10:15:27Z | 2026-04-30T04:38:51Z | 2026-04-26T05:25:40Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | dhltks-hub | 277,708,697 | U_kgDOEI1_mQ | User | false |
openclaw/openclaw | 4,294,906,990 | I_kwDOQb6kR87__xRu | 69,295 | https://github.com/openclaw/openclaw/issues/69295 | https://api.github.com/repos/openclaw/openclaw/issues/69295 | [Feature]: `doctor.memory.status`: expose per-agent `{dirty, files, chunks}` counts | ### Summary
Extend the `doctor.memory.status` RPC response with a `perAgent` array so operator UIs can render per-agent memory state (dirty flag, file count, chunk count) without falling back to the `memory status --json` CLI, which currently hangs on v2026.4.15+.
### Problem to solve
`doctor.memory.status` returns ... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-20T10:17:33Z | 2026-04-29T04:36:27Z | 2026-04-26T12:05:34Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | mehulbhardwaj | 16,291,665 | MDQ6VXNlcjE2MjkxNjY1 | User | false |
openclaw/openclaw | 4,294,915,247 | I_kwDOQb6kR87__zSv | 69,296 | https://github.com/openclaw/openclaw/issues/69296 | https://api.github.com/repos/openclaw/openclaw/issues/69296 | [Bug]: openclaw acp client hangs on slash commands and normal prompts in 2026.4.15 | ### Bug type
Regression (worked before, now fails)
### Beta release blocker
No
### Summary
`openclaw acp client` is broken in OpenClaw 2026.4.15. Repro:
1. run `openclaw acp client -v --server-verbose`
2. wait for prompt
3. enter either `/debug`, `/context`, or even a normal prompt like `hello`
4. client hangs for... | closed | completed | true | 1 | [
"bug",
"regression"
] | [] | 2026-04-20T10:18:55Z | 2026-04-30T04:38:50Z | 2026-04-26T05:25:44Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | mwaeckerlin | 4,056,726 | MDQ6VXNlcjQwNTY3MjY= | User | false |
openclaw/openclaw | 4,294,939,319 | I_kwDOQb6kR87__5K3 | 69,298 | https://github.com/openclaw/openclaw/issues/69298 | https://api.github.com/repos/openclaw/openclaw/issues/69298 | [Bug]: Codex harness usage inflates session context accounting by double-counting cached input | ## Bug type
Behavior bug (incorrect output/state without crash)
## Summary
This looks specifically related to the `codex` embedded harness path, not just `/status` formatting.
For Codex harness sessions, OpenClaw appears to record usage in a shape where:
- `input` already represents the full prompt token count
- `... | closed | completed | true | 2 | [] | [] | 2026-04-20T10:23:23Z | 2026-04-29T04:33:30Z | 2026-04-26T23:28:47Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | richardmqq | 14,178,267 | MDQ6VXNlcjE0MTc4MjY3 | User | false |
openclaw/openclaw | 4,294,943,656 | I_kwDOQb6kR87__6Oo | 69,299 | https://github.com/openclaw/openclaw/issues/69299 | https://api.github.com/repos/openclaw/openclaw/issues/69299 | [Feature]: Register `skills.uninstall` RPC (mirror of `skills.install`) | ### Summary
Add a `skills.uninstall` RPC method symmetric to the existing `skills.install`, so operator UIs can remove a skill through the same scope-gated, audited transport they use to install one.
### Problem to solve
`skills.install` landed in v2026.4.15 and works well for installing ClawHub skills from any ope... | closed | not_planned | true | 1 | [
"enhancement"
] | [] | 2026-04-20T10:24:13Z | 2026-04-29T04:37:29Z | 2026-04-26T10:41:30Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | mehulbhardwaj | 16,291,665 | MDQ6VXNlcjE2MjkxNjY1 | User | false |
openclaw/openclaw | 4,295,001,597 | I_kwDOQb6kR88AAAABAACF_Q | 69,300 | https://github.com/openclaw/openclaw/issues/69300 | https://api.github.com/repos/openclaw/openclaw/issues/69300 | [Bug]: Agent harness compaction short-circuits OpenClaw memory flush and session bookkeeping, leaving session state unbounded | ## Bug type
Behavior bug (incorrect state / unbounded session growth without crash)
## Summary
When an agent harness (e.g. the `codex` embedded harness) provides its own `compact` handler, OpenClaw hands compaction off to the harness and **does nothing else on its own side**. The harness-side compaction is by design... | closed | completed | true | 1 | [] | [] | 2026-04-20T10:34:53Z | 2026-04-28T04:49:26Z | 2026-04-25T19:07:08Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | richardmqq | 14,178,267 | MDQ6VXNlcjE0MTc4MjY3 | User | false |
openclaw/openclaw | 4,295,036,656 | I_kwDOQb6kR88AAAABAAEO8A | 69,301 | https://github.com/openclaw/openclaw/issues/69301 | https://api.github.com/repos/openclaw/openclaw/issues/69301 | Telegram direct session survives /new and /reset with persisted model/provider/auth override | ## Summary
A Telegram direct-chat session can remain pinned to a non-default model/provider even after `/new` and `/reset`, because the persisted session record keeps override fields instead of being fully cleared.
## Observed behavior
Global default model was:
- `lmstudio/qwen/qwen3.6-35b-a3b`
But the Telegram direc... | closed | completed | true | 1 | [] | [] | 2026-04-20T10:41:22Z | 2026-04-23T04:22:06Z | 2026-04-21T02:36:18Z | NONE | null | 20260423T060020Z | 2026-04-23T06:00:20Z | JuniorAgentOC | 275,673,776 | U_kgDOEG5ysA | User | false |
openclaw/openclaw | 4,295,112,826 | I_kwDOQb6kR88AAAABAAI4eg | 69,303 | https://github.com/openclaw/openclaw/issues/69303 | https://api.github.com/repos/openclaw/openclaw/issues/69303 | [Bug] Inbound dedupe released on error path can replay same message_id after provider failure | ### Bug type
Behavior bug (incorrect output/state without crash)
### Beta release blocker
No
### Summary
When an inbound message is claimed by the gateway dedupe path, then the turn later fails due to provider/auth/runtime errors, OpenClaw releases the in-flight dedupe key without committing a durable dedupe recor... | closed | completed | true | 4 | [] | [] | 2026-04-20T10:55:00Z | 2026-04-29T04:34:04Z | 2026-04-26T21:31:12Z | NONE | null | 20260429T060025Z | 2026-04-29T06:00:25Z | NikolaFC | 54,186,359 | MDQ6VXNlcjU0MTg2MzU5 | User | false |
openclaw/openclaw | 4,295,207,160 | I_kwDOQb6kR88AAAABAAOo-A | 69,306 | https://github.com/openclaw/openclaw/issues/69306 | https://api.github.com/repos/openclaw/openclaw/issues/69306 | [Feature]: Control UI: show clear message dates or day separators in chat view | ### Summary
Show clear message dates in the Control UI chat view
### Problem to solve
In Telegram, the date and time of messages are easy to see in the chat itself. In the OpenClaw Control UI, it is hard to tell on which date a specific message was sent when reviewing a conversation.
The current UI shows compact me... | closed | completed | true | 1 | [
"enhancement"
] | [] | 2026-04-20T11:12:11Z | 2026-04-30T04:38:33Z | 2026-04-26T05:44:02Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | DCS-Rob | 260,031,526 | U_kgDOD3_EJg | User | false |
openclaw/openclaw | 4,295,231,744 | I_kwDOQb6kR88AAAABAAQJAA | 69,307 | https://github.com/openclaw/openclaw/issues/69307 | https://api.github.com/repos/openclaw/openclaw/issues/69307 | Feature request: before_tts modifying hook for per-message voice routing | ## Summary
Add a new plugin hook (`before_tts` or similar) that fires **after** the LLM generates a reply but **before** `parseTtsDirectives()` and `maybeApplyTtsToPayload()` run. This hook should allow plugins to **modify the reply text** — specifically to inject `[[tts:voice=...]]` directives dynamically based on ex... | open | null | false | 1 | [] | [] | 2026-04-20T11:16:57Z | 2026-05-02T16:55:36Z | null | NONE | null | 20260502T180029Z | 2026-05-02T18:00:29Z | IGI30 | 10,922,046 | MDQ6VXNlcjEwOTIyMDQ2 | User | false |
openclaw/openclaw | 4,294,732,321 | I_kwDOQb6kR87__Goh | 69,286 | https://github.com/openclaw/openclaw/issues/69286 | https://api.github.com/repos/openclaw/openclaw/issues/69286 | Bug: session totalTokens not reset after compaction causes infinite safeguard loop | # Bug Report: Session Token Count Not Reset After Compaction
## Summary
After compaction clears messages, `session.totalTokens` is not reset to 0, causing the compaction safeguard to trigger repeatedly in an infinite loop.
## Environment
- OpenClaw Version: 2026.4.15 (041266a)
- OS: macOS (Darwin 25.3.0 arm64)
- Node... | closed | not_planned | true | 3 | [] | [] | 2026-04-20T09:49:00Z | 2026-04-30T04:38:52Z | 2026-04-26T05:25:31Z | NONE | null | 20260430T060018Z | 2026-04-30T06:00:18Z | RrcioElise | 47,847,992 | MDQ6VXNlcjQ3ODQ3OTky | User | false |
openclaw/openclaw | 4,295,160,026 | I_kwDOQb6kR88AAAABAALw2g | 69,304 | https://github.com/openclaw/openclaw/issues/69304 | https://api.github.com/repos/openclaw/openclaw/issues/69304 | Telegram inbound breaks after gateway restart/update with persistent getUpdates 409 self-conflict | ## Summary
After updating OpenClaw and restarting the gateway, Telegram inbound for my OpenClaw bot (`@Ocsambot`) stopped working. Outbound Telegram sends still worked, but inbound messages never reached the agent.
The key detail is that this does **not** appear to be another machine or another bot token using `getUpd... | closed | completed | false | 5 | [] | [] | 2026-04-20T11:03:41Z | 2026-04-25T08:54:20Z | 2026-04-25T08:53:51Z | NONE | null | 20260425T090015Z | 2026-04-25T09:00:15Z | NukerX | 12,994,439 | MDQ6VXNlcjEyOTk0NDM5 | User | false |
openclaw/openclaw | 4,294,836,266 | I_kwDOQb6kR87__gAq | 69,290 | https://github.com/openclaw/openclaw/issues/69290 | https://api.github.com/repos/openclaw/openclaw/issues/69290 | [Bug]: /new and /reset reply "ACP session reset in place" for non-ACP sessions, misleading users and agents | ## Summary
`/new` and `/reset` in a **non-ACP session** (regular native Telegram DM session) incorrectly reply with `✅ ACP session reset in place.`, falsely signaling to the user that their session is ACP-bound. This causes users (and downstream agents) to make wrong assumptions about session capabilities — e.g. belie... | closed | not_planned | true | 4 | [] | [] | 2026-04-20T10:06:30Z | 2026-04-28T04:49:27Z | 2026-04-25T19:07:01Z | NONE | null | 20260428T060020Z | 2026-04-28T06:00:20Z | richardmqq | 14,178,267 | MDQ6VXNlcjE0MTc4MjY3 | User | false |
openclaw/openclaw | 4,294,248,677 | I_kwDOQb6kR87_9Qjl | 69,269 | https://github.com/openclaw/openclaw/issues/69269 | https://api.github.com/repos/openclaw/openclaw/issues/69269 | Bug: compaction/reset paths can break session invariants (messageProvider + token accounting) | ## Problem
Compaction/reset paths can break session invariants in two related ways:
1. hook contexts can lose or mis-normalize `messageProvider`
2. successful compaction can leave stale token/cost accounting behind
In practice this causes downstream consumers to see one real conversation as multiple logical sessions... | open | null | false | 3 | [] | [] | 2026-04-20T08:30:48Z | 2026-05-02T14:22:21Z | null | NONE | null | 20260502T150030Z | 2026-05-02T15:00:30Z | de1tydev | 7,804,799 | MDQ6VXNlcjc4MDQ3OTk= | User | false |
openclaw/openclaw | 4,295,074,849 | I_kwDOQb6kR88AAAABAAGkIQ | 69,302 | https://github.com/openclaw/openclaw/issues/69302 | https://api.github.com/repos/openclaw/openclaw/issues/69302 | Image tool fails with 'Unknown model' for custom OpenAI-compat providers | ## Bug Report
### Summary
The `image` tool fails with `Unknown model: <provider>/<model-id>` when `agents.defaults.imageModel` is set to a model on a custom OpenAI-compatible provider, even though the same model works correctly as the chat model.
### Steps to Reproduce
1. Configure a custom provider with `api: "opena... | closed | completed | true | 3 | [] | [] | 2026-04-20T10:48:22Z | 2026-04-25T04:15:39Z | 2026-04-22T19:32:31Z | NONE | null | 20260425T060021Z | 2026-04-25T06:00:21Z | IoanaApetrei | 10,196,026 | MDQ6VXNlcjEwMTk2MDI2 | User | false |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.